Adaptive brightness compensation driving method for low-glare display assembly

Abstract

The application discloses a low-glare display assembly adaptive brightness compensation driving method, relates to the technical field of vehicle-mounted display, and comprises the following contents: a main processor obtains a bounding box reference coordinate and a light spot motion vector according to a vehicle-mounted light sensing array, an inertial measurement unit and a curved screen three-dimensional topological normal matrix, and writes the micro parameter package of a vertical blanking area. A display driver analyzes the micro parameter package, combines a cross-domain lag time constant and a physical haze constant of an anti-glare coating, deduces a real mapping bounding box coordinate, loads a one-dimensional point spread function filter core, and then generates an asymmetric reverse suppression driving amount, which is combined with a user interface driving signal to drive a panel array. The main processor generates a closed-loop contrast residual according to an ambient light sensor array and executes a state rollback. The method realizes the collaborative connection of parameter domain transmission, near-end compensation and closed-loop correction, and improves the continuity and stability of display compensation in a reflection light scene.

Description

Technical Field

[0001] This invention relates to the field of automotive display technology, specifically to an adaptive brightness compensation driving method for low-glare display components. Background Technology

[0002] With the advent of large-size, high-curvature central control screens in smart cockpits, extreme driving environments, such as direct midday sunlight or driving along tree-lined avenues with varying light levels, often require an anti-glare Ag micro-nano coating with a certain degree of physical haze to prevent specular reflection caused by strong external light. While AG resists strong light reflection, strong light can also create a dense, rapidly shifting white haze on the screen surface, further impacting the contrast of the underlying interface. This results in a grayish appearance in dark areas, a significant drop in image quality, and makes it difficult to read important instruments and warning parameters. Therefore, achieving high dynamic range and real-time adaptive local contrast compensation for anti-glare screens under extreme dynamic lighting conditions is a pressing engineering challenge that needs to be overcome.

[0003] To prevent whitewashing, current technologies mostly employ localized high-brightness overload compensation based on ambient light perception. The SID 3 paper, "Whitewash Compensation Algorithm for In-Vehicle MINI-LED Displays," discloses a workflow for an anti-glare compensation system. Existing anti-glare compensation systems require a powerful cockpit domain controller (SOC) and an underlying TCON to work together. During operation, the SOC uses an in-vehicle ambient light sensor to frequently collect external light source illuminance and incident angle vectors. Then, the GPU in the SOC treats the underlying panel as transparent glass and calculates the whitewash distortion mapping area affected by strong light in a two-dimensional coordinate system. Next, the SOC obtains a full-screen brightness compensation mask proportional to the screen's physical resolution in a full-resolution (e.g., 4K) rendering pipeline based on a two-dimensional point spread function (2D-PSF). In this mask, pixels in the whitewash center area are boosted to a strong positive backlight drive gain (e.g., to 2000 nits), and symmetrical reverse blackening voltages are applied to the boundary pixels near the whitewash area to attempt to establish an isolation zone. Finally, this heavy-load video stream, combining the original UI content and the compensation mask, is passed through frame by frame to the underlying TCON via a high-speed serial video bus such as FPD-link at a fixed frame rate. After receiving and completing the digital-to-analog conversion, the TCON directly drives the underlying physical light-emitting array to output.

[0004] While the global rendering pass-through technology described above can slightly restore contrast in static tests, under real-world, complex, high-frequency dynamic driving conditions, the underlying design suffers a fatal double failure in both physical optics and logical timing. Firstly, from a physical optics perspective, existing compensation algorithms ignore the fact that the outermost Ag coating is also an optical low-pass filter. When a pixel emits extreme high-brightness to combat whitening, these intense photons pass directly through the physically hazy AG coating, resulting in strong "forward secondary volume scattering." This lateral overflow of diffuse photons caused by extreme compensation directly attacks the dark areas that should be weakened, creating extremely glaring internal blooming at the edges of the compensation area, thus turning anti-glare into secondary light pollution that degrades image quality.

[0005] More critically, in terms of heterogeneous communication logic, the traditional structure is trapped in a deadlock between computing power and latency. When the SOC performs heavy full-screen rendering, there is an insurmountable physical transmission latency of 30 to 50 milliseconds when the video bus is queuing to transmit the video to the TCON. When the vehicle is traveling at high speed and the strong light spot sweeps across at an angular velocity of 60° / s, the compensation mask issued by the SOC has expired by the time it reaches the TCON. The "symmetrical blackout dead zone" surrounding the light spot is violently impacted on the normal, glare-free UI interface after the light spot has left, causing high-frequency flickering and extremely blinding spatiotemporal misalignment of the black spot's trailing shadow. If the massive two-dimensional spatial deconvolution algorithm is directly applied to the latency-free terminal TCON, the multi-line video caching requirements at 4K resolution will instantly exceed the less than 128KB srAM physical limit of the automotive-grade TCON, directly causing memory overflow and system crash.

[0006] Therefore, the technical problem to be solved by this invention is: under the dual deadlock of absolute latency and underlying memory limit in cross-domain communication, how to eliminate the secondary diffuse halo and fatal spatiotemporal misalignment black spot trailing caused by local brightness extreme value compensation under the dynamic strong light sweeping condition of large curvature anti-glare screen. Summary of the Invention

[0007] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an adaptive brightness compensation driving method for low-glare display components. The display driver parses a micro-parameter packet, combines the cross-domain hysteresis time constant and the physical haze constant of the anti-glare coating, deduces the real-mapped bounding box coordinates, and loads a one-dimensional point spread function filter kernel to generate an asymmetric reverse suppression driving quantity. This quantity is then synthesized with the user interface driving signal to drive the panel array. The main processor generates a closed-loop contrast residual based on the ambient light sensor array and performs state rollback. This achieves coordinated integration of parameter domain transmission, near-end compensation, and closed-loop correction, improving the continuity and stability of display compensation in reflective scenarios; and solves the technical problems described in the background art.

[0008] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: An adaptive brightness compensation driving method for low-glare display components includes: the main processor calculating the bounding box reference coordinates and the light spot motion vector based on the ambient light field vector output by the vehicle-mounted light sensor array, the pose data of the inertial measurement unit, and the three-dimensional topological normal matrix of the curved screen, and encoding them into a micro parameter packet embedded in the vertical blanking area; the display driver parsing the micro parameter packet, deriving the real mapping bounding box coordinates based on the light spot motion vector and the cross-domain communication lag time constant, and loading a one-dimensional point spread function filter kernel based on the physical haze constant of the anti-glare coating; The display driver generates an asymmetric inverse suppression drive quantity based on the real mapped bounding box coordinates, the one-dimensional point spread function filter kernel, and the direction of the spot motion vector; the display driver then combines the asymmetric inverse suppression drive quantity with the UI drive signal to drive the physical panel array.

[0009] Furthermore, the main processor determines the ambient light field vector based on the sampling values, installation direction, and sensitivity calibration coefficients of the vehicle-mounted optical sensor array, and calculates the bounding box reference coordinates by combining the three-dimensional topological normal matrix of the curved screen; when the inertial measurement unit data is abnormal, the cockpit DMS infrared camera is called to extract the displacement of the driver's corneal reflected light spot and map it into a light spot motion vector. Data anomalies include data packet loss and timestamp confidence levels below the safety threshold.

[0010] Furthermore, the micro parameter packet includes a frame index, bounding box reference coordinates, spot motion vector, time decay weight constant, degradation flag, high priority interrupt flag, and cyclic redundancy check field. The main processor writes the micro parameter packet into the vertical blanking area and sends it to the display driver through the underlying video transmission link, where the display driver latches and parses it during the local refresh cycle.

[0011] Furthermore, in the blind-end coordinate system, the display driver performs forward extrapolation based on the bounding box reference coordinates, the spot motion vector, the time decay weight constant, and the cross-domain communication lag time constant, and incorporates the packet writing delay, parsing delay, line scan delay, and panel electro-optic response delay into the cross-domain communication lag time constant to obtain the true mapped bounding box coordinates.

[0012] Furthermore, the display driver reads the physical haze constant of the anti-glare coating from the read-only configuration area, and determines the direction of the spot motion vector based on the physical haze constant; Select the corresponding one-dimensional point spread function filter kernel in the internal static cache, and complete the loading only when the two-dimensional diagonal step error of the selected filter kernel is not greater than the optical homogenization tolerance calculated from the physical haze constant.

[0013] Furthermore, the display driver forms an outer band based on the boundary of the real-mapped bounding box coordinates, and projects a one-dimensional point spread function filter kernel along a selected axis onto the outer band. It generates a boundary reverse suppression driving amount based on the normal distance from the pixel position to the nearest boundary, wherein the boundary reverse suppression driving amount only acts on the panel pixel array corresponding to the outer band.

[0014] Furthermore, the display driver uses the center of the true mapped bounding box coordinates and the direction of the spot motion vector as the basis for determining the boundary orientation. The boundary where the projection result is located on the forward side is defined as the leading edge side, and the boundary where the projection result is located on the backward side is defined as the trailing edge side. The driving amount is suppressed by the boundary with the first intensity on the leading edge side and attenuated to the second intensity on the trailing edge side according to the trailing relaxation coefficient.

[0015] Furthermore, the main processor performs frame synchronization differential on the gravity acceleration vector output by the inertial measurement unit to obtain the chassis jerk. When the chassis jerk reaches the fuse threshold, a high-priority interrupt flag is written in the next vertical blanking region. After reading the high-priority interrupt flag, the display driver stops the extrapolation update of the real mapped bounding box coordinates and resets the trailing relaxation coefficient to one.

[0016] Furthermore, the display driver synthesizes the asymmetric reverse suppression drive quantity and the UI drive signal within the horizontal and vertical synchronization timing to generate a pulse width modulation duty cycle and form a corresponding source gate voltage pulse, which is then written to the physical panel array; the main processor reads the brightness of multiple measurement points of the ambient light sensor array, calculates the equivalent brightness of the mixed light inside the vehicle, and compares it with the brightness of the target interface to obtain the closed-loop contrast residual.

[0017] Furthermore, the main processor uses the closed-loop contrast residual as an observation, performs Kalman filtering recursion on the time decay weight constant, and writes the updated time decay weight constant into the subsequent micro parameter package; when the magnitude of the light spot motion vector of multiple consecutive frames is lower than the static threshold, the display driver turns off asynchronous extrapolation and smoothly increases the trailing relaxation coefficient to one.

[0018] (III) Beneficial Effects This invention provides an adaptive brightness compensation driving method for low-glare display components, which has the following beneficial effects: The main processor integrates the vehicle-mounted light sensor array, inertial measurement unit, and curved screen's 3D topological normal matrix into a single processing chain, generating bounding box reference coordinates and light spot motion vectors. The compensation basis directly represents the screen's reflection position and sliding direction, preventing the display driver from receiving control objects that deviate from the actual light spot. The bounding box reference coordinates and light spot motion vectors are written into a micro-parameter packet and embedded in the vertical blanking region. This transforms the transmission of the entire screen brightness map into parameter domain transmission, ensuring that data interaction between the main processor and the display driver revolves around the required objects. This reduces the occupation of the transmission chain by irrelevant pixel content, emphasizing the innovation of this solution in transmission organization and processing hierarchy division.

[0019] The display driver deduces the actual bounding box coordinates based on the light spot motion vector, and loads a one-dimensional point diffusion function filter kernel based on the physical haze constant of the anti-glare coating. Timing extrapolation, material properties, and boundary expansion are completed in a coordinated manner to ensure the continuous flow of the driving object's steps. This is also a composite logic different from digital compensation.

[0020] The asymmetric inverse suppression driving quantity is generated by using the real-mapped bounding box coordinates and a one-dimensional point spread function filter kernel. The spot motion vector is divided into boundary suppression according to the leading edge and trailing edge. The directionality and hierarchy of the boundary suppression are allocated according to the spot slip law. The asymmetric inverse suppression driving quantity is combined with the user interface driving signal to drive the physical panel array. The ambient light sensor array is used to generate a closed-loop contrast residual, and the time decay weight constant is corrected. When the spot motion vector is stationary, it is asynchronously extrapolated back, thus connecting the dynamic compensation, feedback correction, and state switching in a closed loop. Attached Figure Description

[0021] Figure 1 This is a diagram showing the overall architecture of the in-vehicle low-glare display compensation system of the present invention. Figure 2 This is a flowchart of the multi-source trigger acquisition, footprint bounding box calculation and spot motion vector generation of the present invention; Figure 3 This is a diagram illustrating the vertical blanking parameter packet structure and embedded transmission timing of the present invention. Figure 4 This is a block diagram of the local extrapolation and physical haze constant gating of the display driver in this invention; Figure 5 This is a schematic diagram of the boundary reverse suppression base field and the leading edge strong suppression and tail relaxation release of the present invention. Figure 6 This is a timing diagram for the switching of chassis accelerometer fuse protection in this invention; Figure 7 This is a closed-loop diagram of the compensation-driven synthesis, light mixing feedback correction, and static state rollback of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figures 1-7 This invention provides an adaptive brightness compensation driving method for low-glare display components. The starting point for compensation in dynamic light fields for automotive low-glare display components is arranged. The starting point is not to generate a full-screen brightness mask first, but to compress the external light stimuli that cause whitening distortion into kinematic parameters that can be transmitted, predicted, and further developed in the display driver.

[0024] To address this, step one involves the main processor coordinating the vehicle-mounted light sensor array, inertial measurement unit, cockpit DMS infrared camera, and underlying video link. It first converges the ambient incident light, curved screen geometry, and vehicle attitude to the same screen coordinate system. Then, it writes the bounding box reference coordinates and the light spot motion vector into the vertical blanking region, allowing subsequent steps to directly receive the parameter stream rather than the entire screen compensation map. From a mechanistic perspective, side-illuminated sunlight, curtain wall reflection, under-bridge light / dark switching, and vehicle pitch all alter the incident direction. The display cover and anti-glare layer then convert this change into a whitening distortion footprint sliding along the screen surface. Step one does not analyze interface icons and color blocks; it only seeks the position and displacement trend of the footprint. Therefore, the output object is always locked to the bounding box reference coordinates and the light spot motion vector.

[0025] Step 1: On the main processor side, the changes in the external multi-source light field and the vehicle attitude are converged into bounding box reference coordinates and light spot motion vectors decoupled from the image content, and sent to the display driver in the form of a lightweight parameter package that can be embedded in the vertical blanking area.

[0026] In-vehicle displays operate within a constantly changing external light field and vehicle posture. If the main processor still generates full-resolution luminance masks frame by frame and then sends them down in their entirety via the underlying video link, the compensation process would be slowed down by pixel movement and cross-domain queuing. Therefore, step one first addresses the physical chain of "where the light comes from, where it hits the screen, and in which direction it slides," rather than first addressing "which image is on the interface." The problem is not simply identifying glare, but rather separating glare from the image domain to the kinematic parameter domain, leaving a directly callable input object for the asynchronous extrapolation in step two.

[0027] The unified execution entity is the main processor. The vehicle-mounted light sensor array is preferably positioned along the upper edge of the dashboard, the interior surface of the A-pillar, or the back side of the upper frame of the curved screen, and its installation coordinates and orientation are written during vehicle calibration. The vehicle-mounted light sensor array preferably includes at least four non-coplanar directional photosensitive units, each of which pre-stores the installation direction vector and sensitivity calibration coefficients. The inertial measurement unit continuously transmits pitch, yaw, roll, and linear acceleration data from the chassis domain via a CAN bus or FlexRay bus. The three-dimensional topological normal matrix of the curved screen is generated offline and stored based on the assembly model of the inner surface of the cover glass and the light-emitting surface of the light-emitting panel during assembly manufacturing. With this arrangement, the input object, the processing entity, and the output destination are always within the same single chain.

[0028] The main processor first performs dark current subtraction, saturation culling, and display frame time base alignment on the original sampled values ​​of each channel of the vehicle-mounted light sensor array, and then reconstructs the ambient light field vector according to the above formula; when the magnitude of the ambient light field vector exceeds the first illuminance threshold... Afterwards, the main processor enters the glare preprocessing branch; then it maps the ambient light field vector, the three-dimensional topological normal matrix of the curved screen, and the pose data of the inertial measurement unit to a unified screen coordinate system to obtain the white distortion footprint bounding box.

[0029] Next, the derivative of the bounding box coordinates after pose compensation is obtained to extract the spot motion vector of the stripped image pixels. If the inertial measurement unit (IMU) loses packets, fails verification, or the timestamp confidence is insufficient, the spot motion vector acquisition branch is switched to the cockpit DMS infrared camera. The degraded motion vector is obtained by substituting the driver's corneal reflection spot displacement, head pose stripping, and eye-screen mapping back, while the bounding box reference coordinates are still obtained from the ambient light field vector and the three-dimensional topological normal matrix of the curved screen. Finally, the bounding box reference coordinates, spot motion vector, time confidence, time decay weight constant, degrade flag, high-priority interrupt flag, and cyclic redundancy check are encapsulated into a vertical blanking parameter package and embedded into the underlying video link. The screen coordinates and time stamps output by the previous step are inherited by the next step as is, so step two does not require reviewing the original sensor.

[0030] The vehicle-mounted optical sensor array does not directly send the raw photosensitive level into the compensation chain. Instead, it first reconstructs the ambient light field vector within the main processor. Similarly, the inertial measurement unit does not input all chassis quantities as is. Instead, it only retains pitch, yaw, and tilt quantities directly related to screen attitude changes, along with their corresponding time stamps. This processing method organizes data around objects that alter the positional relationships of the white distortion footprint, preventing quantities unrelated to the screen from being mixed into the main chain.

[0031] For example, when a vehicle enters a side-sunlit section from an underground parking garage ramp, the light-sensing units along the upper edge of the dashboard experience sudden brightness increases. Upon receiving this trigger, the main processor does not immediately rewrite the entire interface; instead, it first locks the current ambient light field vector and the most recent inertial measurement unit pose sequence. The driver sees one side of the screen begin to turn white, while the main processor internally unifies the input to the same frame index, awaiting the next geometric calculation. If the curved screen is located in the center console area, the light-sensing units can also be arranged on the lower surface of the center console sunshade. As long as the installation coordinates and the screen coordinate system have a definite mapping, the subsequent process remains consistent.

[0032] In specific implementation, the main processor preferably uses piecewise cubic Hermite interpolation to resample the sampling times of the vehicle-mounted photosensitive array to the display frame time base, and then uses five-point differential to obtain the attitude change trend from the pose sequence of the inertial measurement unit. The former maintains position and slope continuity, and will not stretch local photosensitive peaks into long tails; the latter retains the turning direction, which is convenient for subsequent determination of the sign of the light spot motion vector. After the time base convergence is completed, each frame has a unique frame index, a unified time label, and a unified screen coordinate reference plane, and processing method B directly inherits this result. Furthermore, the triggering conditions are clear, the time base is single, and the photosensitive unit arrangement under different vehicle models retains an equivalent implementation path.

[0033] Curved screens are not flat panels with a single normal; the left, center, and right sides respond differently to the same incident direction. Therefore, the main processor does not use a uniform incident angle for the entire screen. Instead, it reads the local normals in the three-dimensional topological normal matrix of the curved screen one by one, and then projects the surface sampling points that meet the whitening threshold onto the screen coordinate system to form a footprint set. The minimum bounding box is then extracted from the footprint set. The reference coordinates of the bounding box obtained in this way directly preserve the coupling relationship between curvature, assembly posture, and incident direction.

[0034] Where: the set of footprints Indicates the first A set of screen footprint points that satisfy the local whitewash entry condition; its elements are two-dimensional coordinate points in the screen coordinate system. (Bounding box coordinates) Indicates the first Frames are composed of a set of footprints The extracted whitening distortion footprint bounding box; the preferred value is a combination of the coordinates of the top left and bottom right corners in the screen coordinate system, or equivalently, a rectangular boundary parameter composed of the left, top, right, and bottom boundaries. (Screen coordinates) Indicates the first The surface sampling point at the first The coordinates of the frame mapped to the screen coordinate system; the values ​​are located within the effective display area, preferably a two-dimensional column vector or a two-dimensional coordinate pair. Surface sampling points. The first discretized part of the curved screen surface represents the... Three-dimensional sampling points; values ​​are located on the surface of the curved screen cover, the light-emitting surface of the light-emitting panel, or a predetermined reference surface, and are generated by offline geometric discretization. Screen projection operator. : This represents the projection operator that maps sampling points on a 3D surface to the screen coordinate system; its function is to map sampling points on the curved surface of the screen space. Converted into two-dimensional screen coordinates that can be directly processed by the display driver. Its specific implementation can be completed based on the screen reference surface geometric extrinsic parameters and screen pixel grid definition after vehicle calibration.

[0035] Local normal Indicates the first Surface sampling points The corresponding unit normal vector; its magnitude is 1, derived from the 3D topological normal matrix of the curved screen or the discrete geometric model of the curved surface. Ambient light field vector. Indicates the first The frame is the ambient light field vector reconstructed from the vehicle-mounted light sensor array; the magnitude is positive and originates from the reconstruction results output by the main processor from the multi-directional light sensor units. Magnitude : Represents the ambient light field vector The Euclidean norm; its function is to normalize the ambient light field vector, making the judgment formula... It primarily reflects the degree of directional consistency between the local normal and the incident direction, rather than being directly affected by the amplification of the absolute magnitude of light intensity; its value is positive.

[0036] Localized white threshold : indicates the first Each surface sampling point enters the footprint set The threshold for judgment; the preferred value is located in the open interval. This is written from the vehicle calibration, factory calibration, or regional calibration table. Enclosure operator. : This represents the operator that extracts the minimum bounding rectangle of the input point set in the screen coordinate system; its function is to extract the discrete footprint set. The compression is performed into a rectangular bounding box representation that can be efficiently transmitted and processed in subsequent links; its output is the smallest rectangular boundary that can completely enclose all points of the footprint set. Sampling point index. : Represents the number of the sampling point on the discrete surface of the curved screen; its function is to traverse all sampling positions in the three-dimensional topological discrete model of the curved screen; its value is a positive integer. Frame Index : Indicates the time sequence number of the display frame or processing frame; its function is to distinguish the ambient light field, footprint set and bounding box result on different time slices; its value increases with the display time base or unified processing time base.

[0037] The main processor uses the above formula to filter out those that meet the threshold for localized whitening. The sampling points are then used to generate bounding box coordinates. Due to the localized whitening threshold During the vehicle assembly stage, the directional characteristics of the cover glass surface have been incorporated, ensuring that the edge areas with rapid curvature will not be excessively expanded due to occasional oblique radiation.

[0038] For example, when a vehicle turns right through a glass curtain wall section, the upper right area of ​​the screen is first included in the footprint set, and then the footprint smoothly slides along the upper middle part of the screen. The bounding box reference coordinates output by the main processor show continuous migration rather than full-screen flickering. If the display assembly is an integrated dual-screen, the three-dimensional topological normal matrix of the curved screen can be stored in left and right segments and then stitched together into the same screen coordinate system, with the formula remaining unchanged.

[0039] When in use, the bounding box coordinates are directly generated from the relationship between the ambient light field vector and the local normal, avoiding full-screen pixel traversal and providing a stable position reference.

[0040] Furthermore, the main processor does not perform optical flow on image pixels. Instead, it directly differentiates the bounding box reference coordinates of consecutive frames after pose compensation, thus obtaining pure motion vectors stripped of image content. The main processor first calculates the time confidence based on the inertial measurement unit integrity flag and timestamp deviation, and then uses the inertial main chain when the time confidence meets the safety threshold. If the inertial measurement unit experiences packet loss, verification failure, or a decrease in timestamp confidence, the main processor calls the cockpit DMS infrared camera to convert the displacement of the corneal reflected light spot into a degraded motion vector through head-motion stripping and eye-screen mapping.

[0041] Finally, the main processor writes all objects into a vertical blanking parameter package and embeds it into the underlying video link, allowing the display driver to directly access it without increasing external bandwidth. This vertical blanking parameter package includes, in addition to the bounding box reference coordinates and spot motion vector, a frame index, temporal confidence, temporal decay weight constant, degradation flag, high-priority interrupt flag, and a cyclic redundancy check field. Where: fusion weight coefficient Indicates time For the main chain motion vector of the inertial measurement unit The acceptance weight; the preferred value is located in the closed interval [0,1]; the basic integrity weight. Indicates time The basic reliability of the inertial measurement unit (IMU) data packets; the value is preferably located in the closed interval [0,1]. When the IMU experiences packet loss, packet corruption, or verification failure, a lower value or zero can be used. Current frame time interval. Indicates time The unified time base time interval between the current frame and the previous frame; the nominal sampling period of the inertial measurement unit. : Indicates the design sampling period or nominal packet interval of the inertial measurement unit; the value is positive and is given by the hardware parameter table or the factory configuration table.

[0042] Maximum permissible time deviation This represents the maximum permissible deviation scale of the actual time interval of the inertial measurement unit relative to the nominal sampling period; its value is positive and is given by the calibration parameter table. Main chain motion vector of the inertial measurement unit. Indicates time Candidate motion vectors are obtained based on pose compensation from the inertial measurement unit; their output is used for weighted fusion with the degraded branch results from the cockpit DMS. Head motion decoupling matrix. : Represents the decoupling matrix that maps head pose changes to the apparent displacement components of the eye region; its value is obtained from vehicle calibration, cockpit DMS geometric model, or offline fitting. Current frame head pose feature vector Indicates time The driver's head position and attitude feature vector estimated by the cockpit DMS infrared camera; the content may include one or more of the following: head position, pitch angle, yaw angle, and roll angle.

[0043] The head pose feature vector of the previous frame : Indicates time The driver's head pose feature vector estimated by the cockpit DMS infrared camera; its function is to... Together, they constitute the pose changes in the head region of two adjacent frames. Finally, the motion vectors are fused. Indicates time Candidate motion vectors of the main chain of the inertial measurement unit Candidate motion vectors of the cockpit DMS degraded branch By fusion weight coefficient The final light spot motion vector obtained by weighting; bounding box center operator This operator converts input bounding box coordinates to bounding box center coordinates; its function is to compress the rectangular boundary position into a single point, thus facilitating rigid body mapping and velocity calculation between frames; its output is preferably two-dimensional center coordinates in the screen coordinate system. Current frame bounding box coordinates. Indicates time The corresponding white-distorted footprint bounding box; its function is to serve as the current position object in solving the main chain motion vector of the inertial measurement unit; its source is the bounding box reference coordinates obtained after extracting the previous footprint set in step one, or the bounding box coordinates updated in the same frame. Previous frame bounding box coordinates. : Indicates time The corresponding whitened distortion footprint bounding box; Inertial pose mapping function : Indicates the time according to the inertial measurement unit at time With time The pose changes between frames are mapped by a function that maps the rigid body at the center of the bounding box in the previous frame to the screen coordinate system in the current frame; internally, this can be achieved by rotational components, translational components, and screen coordinate projection relationships. (Cockpit DMS degraded branch motion vector) Indicates time In the event of inertial measurement unit malfunction, packet loss, verification failure, or insufficient time confidence, candidate motion vectors are obtained inversely from the cockpit DMS infrared camera branch; eye-to-screen coordinate mapping matrix. This represents the mapping matrix that transforms the driver's local eye reference coordinates or corneal reflection displacement coordinates to the screen coordinate system; its values ​​are obtained from vehicle calibration, eye position calibration, or human-machine line-of-sight geometry calibration. Current frame corneal reflection center coordinates. Indicates time The coordinates of the center of the corneal reflection spot extracted by the cockpit DMS infrared camera, or equivalently, the coordinates of the representative point of the corneal reflection after tracking; the source is the result of detection, tracking or optical flow processing of the corneal reflection area within the iris highlight area.

[0044] Coordinates of the corneal reflection center in the previous frame : Indicates time Coordinates of the center of the corneal reflected light spot extracted by the cockpit DMS infrared camera; Time index It represents the time sequence number of the current processing frame or the current moment; its function is to distinguish the confidence, bounding box, head pose and motion vector in different time slices.

[0045] The main processor first uses the pose mapping function Move the center of the bounding box from the previous frame to the current screen coordinate system, then subtract the center of the bounding box from the center of the current frame and divide by the time interval. The inertial motion vector is obtained. The key is to perform pose compensation before differentiation. Therefore, what is obtained is not the apparent drift caused by vehicle body swaying, but the footprint slip trend after accounting for changes in vehicle attitude. For example, when a vehicle continuously passes over speed bumps, the driver can see reflections and vibrations accompanying the vehicle body's undulations, but the inertial motion vector output by the main processor... It still points in the true direction of lateral slip.

[0046] Furthermore, when the inertial measurement unit experiences packet loss, verification failure, or insufficient timestamp confidence, the main processor switches to the cockpit DMS infrared camera to acquire infrared images of the driver's eyes. This camera preferably uses narrowband infrared illumination and a fixed-focus lens. First, it extracts the corneal reflection spot area from the iris's high-brightness region. Then, it performs Horn-Schunck continuous field optical flow, Farnebäck dense optical flow, or pyramidal Lucas-Kanade local optical flow cluster tracking on the continuous frames of the corneal reflection spot area to obtain the corneal reflection displacement. The main processor further estimates the head pose change between the two frames, extracts the common displacement component caused by the head pose from the corneal reflection displacement, and converts it into a degraded motion vector in the screen coordinate system based on a pre-calibrated eye-screen mapping matrix.

[0047] Light spot motion vector : Represents the final output motion vector of step one, including direction and velocity in the screen coordinate system; Time confidence : Indicates the reliability of the time stamp of the inertial measurement unit in the current frame, and the value is located in a closed interval. ;in As a marker of integrity, For the nominal sampling period, Upper limit of allowable deviation; inertial motion vector : Represents the candidate motion vector obtained based on the inertial measurement unit; degraded motion vector : represents the candidate motion vector obtained from corneal reflection displacement through head motion decoupling and eye-screen mapping; where Represents the eye-screen mapping matrix; corneal reflection centroid. Head posture characteristics and head motion decoupling matrix : These represent the centroid coordinates of the corneal reflective spot region after tracking, the head pose estimation result, and the head motion compensation mapping relationship in the two consecutive infrared images; Time decay weighting constant Downgrade indicator and high-priority interrupt flag : These represent the attenuation coefficient used to weight the light spot motion vector during extrapolation in step two, whether the current parameter source is in a degraded branch, and whether a mechanical abrupt change has been detected; where Initialize to 1, and then update in a closed loop as per step four.

[0048] Parameter package Indicates time The entire lightweight parameter packet is written to the vertical blanking region of the underlying video link; the output destination is the parameter receiving and local extrapolation processing chain on the display driver side. (Bounding box coordinates) Indicates time The corresponding bounding box coordinates of the white distortion footprint; the preferred values ​​are a combination of the coordinates of the top-left and bottom-right corners in the screen coordinate system, or equivalently, rectangular boundary parameters composed of the left, top, right, and bottom boundaries. Frame Index Indicates time The frame sequence number identifier corresponding to the parameter packet; the value is generated incrementally according to the unified display time base or the unified processing time base. Cyclic Redundancy Check field. Indicates time The result of the cyclic redundancy check (CRC) on the corresponding parameter packet body; its function is to allow the display driver to quickly determine whether the parameter packet has suffered bit errors, truncation, or field corruption during link transmission, thereby deciding whether to accept the current frame's parameter packet or retain the valid parameters from the previous frame; its value is calculated by the main processor based on the parameter packet body according to a preset check polynomial. Time Index Indicates the current processing frame or the time sequence number under the current unified time base; During the packaging stage, the main processor sets the bounding box coordinates. Light spot motion vector Time confidence Time decay weighting constant Downgrade indicator High-priority interrupt flag The frame index and cyclic redundancy check (CRC) are written into the vertical blanking parameter packet. The vertical blanking parameter packet is preferably placed in the vertical blanking region of the FPD-Link link. The coordinate field uses a fixed-point format to store the top-left and bottom-right corners of the bounding box, the motion field uses a fixed-point format to store the horizontal and vertical components, and the weight field uses a fixed-point format. The degradation flag and the high-priority interrupt flag are each occupied by a separate bit, and cyclic redundancy check is used for rapid acceptance on the display driver side. After receiving the data, the display driver can read the object within its local refresh cycle without accessing the main processor's original sensor cache.

[0049] In use, the first step compresses high-dimensional sensing data into lightweight objects that can be directly consumed by the underlying driver, while maintaining continuous switching between the inertial main chain and the infrared branch chain.

[0050] When using it, due to the bounding box coordinates It directly inherits the geometric relationship between the screen normal matrix and the ambient light field vector, and the light spot motion vector. Furthermore, since the derivatives of consecutive frames after pose compensation are physically consistent, step two can directly deduce the mapping bounding box at the current physical moment. Also, because the vertical hidden line culling parameter package already includes temporal confidence... Time decay weighting constant With demotion sign The display driver can determine whether the current object comes from the inertial main chain or the infrared branch chain when reading parameters, and will... It acts directly on the extrapolation chain, thereby maintaining the continuity of the subsequent processing chain.

[0051] During engineering verification, the continuity of bounding boxes in consecutive frames, the pass status of cyclic redundancy check of vertical blanking parameter packets, and the consistency of the direction of light spot motion vector before and after degradation switching can be used as evaluation indicators; for comparison paths, the pose mapping function can be turned off, and only the directional continuity of the uncompensated derivative and the output of this step in the vehicle body undulation road section can be compared.

[0052] Step 2: The display driver sets the bounding box reference coordinates within the local refresh cycle. and light spot motion vector Extrapolation is performed to the actual bounding box coordinates, based on the physical haze constant. Choose a one-dimensional point spread function filter kernel that satisfies the optical homogenization tolerance.

[0053] Step one has already compressed the ambient light field, vehicle attitude, and degradation information into a miniature parameter packet. If this parameter packet is then sent back to the main processor to be expanded into a two-dimensional compensation map and then transmitted back to the display driver, cross-domain wait will be reintroduced. Therefore, step two shifts the computational focus to the display driver, allowing it to process glare objects only on the local clock, row synchronization, and column drive clock cycles. The sliding of the glare footprint on the display panel is itself a spatial displacement phenomenon directly coupled to the display drive clock cycle. The closer it is to the column drive and gate drive execution surfaces, the easier it is to compress the time difference within the local link.

[0054] On the other hand, step two does not treat the one-dimensional point spread function filter kernel as a simple digital convolution template, but rather uses the physical haze constant of the outermost anti-glare coating of the display screen. This is all included. Because the anti-glare coating itself performs a physical low-pass on the emitted light, if the digital side is still expanded according to two-dimensional high-precision compensation, it will waste the on-chip cache; if the digital side is overly coarsened, it will create a diagonal staircase effect. Step two is to establish a fixed mapping between the two, so that the digital truncation error falls within the range that physical fogging can swallow.

[0055] During each frame's vertical blanking, the display driver receives a tiny parameter packet written by the main processor and parses out the frame index and bounding box reference coordinates. Light spot motion vector Time decay weighting constant The display driver first maps the parameter packet to the blind-end coordinate system based on the local line clock, and then uses the system-predefined cross-domain communication lag time constant. bounding box reference coordinates Step-by-step accumulation yields the true bounding box coordinates of the current physical moment; subsequently, the physical haze constant is read from the fuse region or read-only register region. The physical haze constant The values ​​are converted to a low-pass cutoff frequency and a core tap length, and then the corresponding one-dimensional point spread function filter core is loaded from the on-chip static random access memory. After the coordinate results and core results are closed inside the display driver, step three can directly form an asymmetric reverse suppression drive quantity distribution based on these two results.

[0056] First process the bounding box reference coordinates The problem of resolving the time difference between the displayed driver and the actual mapped bounding box coordinates. The result is the moment the main processor sends the packet, not the moment the local column driver actually lands. Therefore, the display driver first writes the parameter packet to the double-page register cache during the vertical blanking period, and then the timing microcode reads it step by step between row synchronizations. The blind-end coordinate system here refers to the internal grid coordinate system of the display driver, with the upper left corner of the display area as the origin, the horizontal direction as the x-axis and the vertical direction as the y-axis, and the coordinate unit as pixel pitch. Within this coordinate system, the display driver only cares about the pixel grid and time index, and no longer perceives the raw data from the external light sensor array or inertial measurement unit. Therefore, its extrapolation object is always the bounding box reference coordinate. Time decay weighting constant With the light spot motion vector The combination of .

[0057] The fields of the miniature parameter packet are preferably written to the vertical blanking area in the following order: start identifier, frame index, top-left coordinates of the bounding box, bottom-right coordinates of the bounding box, horizontal velocity component, vertical velocity component, time decay weight constant, degradation flag, high-priority interrupt flag, and cyclic redundancy check. If the display driver verification passes, the bounding box reference coordinates are then written to the blanking area. Time decay weighting constant With the light spot motion vector Lock the data pair to the same frame; if the verification fails, retain the data pair that has been confirmed in the previous frame and prohibit updates in the current frame to avoid mismatch of coordinates, weights and directions across frames.

[0058] Where: the coordinates of the actual mapped bounding box : indicates the first Frame in The extrapolated results on each local refresh substep are within the coordinate range of the display valid area; bounding box reference coordinates : Represents the original bounding box coordinates delivered in step one, which is the starting point of the extrapolation; local refresh substep number. : Indicates the number of times the display driver performs an accumulation within a cross-domain communication hysteresis window, with values ​​ranging from 0 to the local accumulation limit. Integers; Local refresh substep duration : Represents the time interval between two blind end coordinate updates within the display driver, and takes a positive value; Spot motion vector : Represents the direction and speed of motion delivered in step one, and is a two-dimensional vector in the screen coordinate system; where the time decay weight constant is... After weighting, it is used as the actual extrapolated speed. Initialize to 1 and update via closed-loop process in step four; Cross-domain communication lag time constant : Represents the calibration time difference between the main processor writing the parameter packet and the display driver actually executing the drive; the value is positive. The delay of writing packets to the main processor For display driver resolution delay, For the center row of the bounding box The corresponding row scan delay, For panel electro-optical response delay; Local accumulation limit : Represents the number of discrete accumulations that can be executed within the hysteresis window of cross-domain communication, and the value is a non-negative integer; Projection clipping operator : This operator clips the extrapolated bounding box to the displayable area. Specifically, it performs upper and lower boundary clamping on both the x and y coordinates. (Bounding box center row coordinates) Indicates the first The frame bounding box center's row coordinates in the vertical direction of the screen coordinate system; its function is to determine the panel scan row position corresponding to the current bounding box center, and to calculate the row scan delay from the start of frame synchronization to the actual scanned position; its value lies within the vertical coordinate range of the display's effective area, and can be determined from the bounding box coordinates. Obtained through the central operator. Time index. Indicates the current processing frame or the time sequence number under the current unified time base; mapping identifier superscript This indicates that the corresponding bounding box coordinates are the actual mapping results after extrapolation via a local refresh substep and projection clipping; line scan delay. : Represents the row coordinates of the bounding box center The corresponding panel scan delay; the value is determined by the row coordinates of the bounding box center. It is determined together with the panel row scan cycle.

[0059] For example, when a vehicle enters a high-side-light road section from the tunnel exit, the main processor delivers the bounding box reference coordinates. The image remains at the screen position corresponding to the exit edge, while the vehicle has already continued moving forward when the display driver receives the signal. At this point, the display driver does not wait for the next frame to re-identify, but immediately performs a local refresh substep after vertical blanking. Continuous accumulation results in the bounding box moving forward along the original sliding direction and realigning with the current reflective position. This transforms cross-domain waiting into local discrete accumulation.

[0060] Further investigation is needed into why the display driver only loads a one-dimensional point spread function filter core and not a two-dimensional core. After extrapolating the coordinates, the display driver does not immediately fetch the fixed core from static random access memory; instead, it first reads the physical haze constant of the outermost anti-glare coating of the display screen. Physical haze constant The haze is measured by an integrating sphere haze meter when the display assembly leaves the factory and written to the fuse area or read-only register area of ​​the display driver.

[0061] In a preferred embodiment, the anti-glare coating utilizes a surface scattering structure formed by silicone resin and spherical silica microparticles. The silica microparticles are used to create a controlled micro-roughness surface, while the silicone resin is used to fix the microparticles and maintain transmittance. After curing, a stable atomization layer is formed on the outer surface of the cover plate. Step two does not re-involve in the coating preparation but directly uses the physical haze constant measured at the factory. If an etched glass anti-glare layer or an organic hardened coating is used instead, as long as the physical haze constant of the same aperture can still be obtained, it is possible to achieve the same result. The gating logic of the display driver still holds. The on-chip static cache of the display driver pre-stores multiple families of one-dimensional point spread function filter kernels. Each family of filter kernels is obtained by integrating, normalizing, and quantizing the two-dimensional point spread function along the main motion axis in the corresponding haze interval.

[0062] Where: Low-pass cutoff frequency : indicates the first The cutoff boundary corresponding to the frame's one-dimensional point spread function filter kernel takes a positive value; the reference spread time... : Represents the minimum diffusion time preset for the baseline panel without a fog layer, and takes a positive value; fog mapping coefficient : Represents the physical haze constant The gain coefficient for diffusion time is taken as a positive value; physical haze constant. : Indicates the measured haze property of the outermost anti-glare coating, and the value is positive; Direction correction factor : Represents the additional influence coefficient of oblique motion on the low-pass boundary, and takes a non-negative value; motion direction angle : Represents the motion vector of the light spot The orientation angle in the screen coordinate system takes the value located at... arrive Between; vertical velocity components : Represents the motion vector of the light spot The component on the vertical axis can take either positive or negative values; the horizontal velocity component... : Represents the motion vector of the light spot The component on the horizontal coordinate can take either positive or negative values; arctangent mapping operator This represents an operator for calculating the complete quadrant orientation angle based on the horizontal and vertical velocity components. Tolerance mapping coefficient. This represents the physical haze constant. Mapped to optical homogenization tolerance The proportionality coefficient; it takes a positive value and is obtained from panel calibration, material calibration or whole-machine optical calibration.

[0063] Optical homogenization tolerance Indicates the first The upper limit of the one-dimensional point spread function filter kernel approximation error that the frame can tolerate under the current physical haze conditions; its function is as a diagonal step error. The constraint boundary is used to determine whether the selected one-dimensional point spread function filter kernel is still within the acceptable range that physical atomization can absorb; its value is positive and is determined by the formula. Sure.

[0064] Diagonal step error Indicates the first The frame-selected one-dimensional point spread function filter kernel generates discrete step error when approximating oblique spread; its function is to characterize the degree of geometric discrete distortion introduced when the one-dimensional kernel approximates two-dimensional oblique spread, and to correlate it with optical homogenization tolerance. The comparison determines whether the current kernel is allowed to load; its value is non-negative. Pi (π) constant. Pi is a constant. The display driver utilizes a low-pass cutoff frequency. The purpose is not to perform spectral analysis, but to select a bandwidth boundary that matches the material scattering for the subsequent one-dimensional point spread function filter kernel; when the physical haze constant... When the glare is high, the anti-glare coating itself already handles stronger outgoing diffuse scattering, so the display driver lowers the low-pass cutoff frequency. The adjustment makes the digital core smoother; when the light spot motion vector... When pointing diagonally, the direction correction factor Make the low-pass cutoff frequency This narrows the range further. Therefore, material scattering can be directly converted into digital gating conditions.

[0065] Further solutions are needed to accommodate a sufficient number of one-dimensional point spread function (DFD) filter kernels within a very small static random access memory (SRAM) while keeping the dimensionality reduction error within the range that physical fogging can cover. The display driver pre-stores multiple one-dimensional DFD filter kernels with odd-numbered tap lengths. Each kernel is stored symmetrically according to its center coefficient and quantized in a fixed-point format. The kernel coefficients are represented using twelve-bit fixed-point, and the sum of the center coefficient and the coefficients on both sides is normalized to one, avoiding overall panel brightness drift after loading. Corresponding kernel families are indexed according to fog level. The number is generated offline by integrating, normalizing, and quantizing the two-dimensional point diffusion function template measured at the factory stage along the main motion axis.

[0066] The display driver operates according to the low-pass cutoff frequency. First select the bandwidth family, then select the tap length, and then decide whether the filter core is loaded along the horizontal or vertical axis; when the motion direction angle When in the oblique region, the display driver preferentially selects the axis containing the principal component to load a one-dimensional point spread function filter kernel, and applies a phase shift to adjacent rows or columns, thereby approximating the oblique expansion without upgrading to a two-dimensional kernel. The display driver only performs operations when... In the nuclear family, the nominal cutoff frequency and the current cutoff frequency are selected. The kernel family loading with the smallest deviation.

[0067] Where: the length of the core tap : indicates the first The length of the one-dimensional point spread function filter kernel loaded into the frame is an odd number greater than or equal to 3; the tap scale factor. : Indicates the scaling factor by which the display driver converts the low-pass cutoff frequency to the tap length; a positive value is taken. Low-pass cutoff frequency : Indicates the obtained cutoff boundary, with a positive value; diagonal step error : Represents the discrete step size formed when the one-dimensional point spread function filter kernel is approximately obliquely expanded; the value is non-negative; pixel pitch. : Represents the center distance per unit pixel of the display panel, and takes a positive value; motion direction angle : Represents the motion vector of the light spot The direction angle, the value is located in arrive Between; 12-bit quantization operator This usually indicates that the real number in parentheses is mapped to a 12-bit fixed-point representation or a discrete value with 12-bit quantization precision. Optical homogenization tolerance : Represents the physical haze constant The derived upper limit of the tolerable discrete error takes a positive value; tolerance mapping coefficient : Represents the physical haze constant To optical homogenization tolerance The conversion factor is positive; physical haze constant. : Represents the measured haze property of the outermost anti-glare coating, with a positive value. Kernel tap coefficient : indicates the first The one-dimensional point spread function filter kernel corresponding to each haze level is at the th... The discrete kernel coefficients at each tap position; after normalization, the values ​​satisfy that the sum of all tap coefficients is 1, and preferably are non-negative values.

[0068] Two-dimensional point diffusion function template cross section : indicates the first The continuous values ​​of the two-dimensional point diffusion function template under each haze level are taken on the main motion axis section; the preferred source is the measured diffusion response of a standard bright line pattern after passing through the cover plate and the outermost anti-glare coating, or obtained by fitting the corresponding optical model. Level index : Indicates the physical haze constant classification or kernel family classification number; the value is determined by the current physical haze constant. The interval is determined. Tap index. : indicates the first The tap position number in a one-dimensional point spread function filter kernel; the value lies within the discrete tap range corresponding to the tap length of that kernel. Continuous spatial coordinates. : Represents a one-dimensional continuous spatial coordinate system unfolded along the main motion axis; its function is to display the coordinates in pixel increments. Within the defined discrete tap intervals, integration is performed over the two-dimensional point spread function template cross section to obtain the continuous energy contribution of the corresponding tap. (Normalized summation index) : Represents the intermediate summation index in the denominator used to traverse all discrete tap intervals; its function is to calculate the summation index of the current tap interval. The integral results of each kernel family over all tap intervals are normalized and summed to ensure that the sum of the generated tap coefficients is 1. Kernel family tap length. : indicates the first The length of the one-dimensional point spread function filter kernel corresponding to each haze level; its function is to limit the tap range of the kernel family for that level and the upper and lower bounds of the normalized summation of the denominator; its value is an odd number greater than or equal to 3, which is determined during the offline kernel design stage.

[0069] Specifically, the display driver does not aim to eliminate diagonal step error. Instead of compressing it to zero, only the diagonal step error is required. Not exceeding the physical haze constant Derived optical homogenization tolerance .

[0070] The digital side allows for some degree of discreteness, as long as this discreteness doesn't create a visually perceptible hard edge after passing through the anti-glare coating. For example, when the setting sun shines obliquely through the passenger-side window onto the curved central control screen, the display driver first maps the actual bounding box coordinates... Locked to the right side of the screen, then selects the low-pass cutoff frequency from static random access memory. The corresponding odd-numbered tap cores are loaded horizontally; due to the slightly upward direction of the reflective footprint, the display driver simultaneously performs a one-pixel phase shift between adjacent rows, resulting in a softly shifted compensation boundary. If the panel manufacturer uses different pixel pitches... For different cover plate materials, only the tap dimensional coefficients need to be recalibrated. And tolerance mapping coefficient The kernel family and load timing in the static random access memory can still be used. Thus, without increasing the cache area, the digital approximation error of the one-dimensional point spread function filter kernel is physically atomized and swallowed up.

[0071] When in use, the display driver outputs the actual mapped bounding box coordinates that have been pushed to the current physical moment within the local refresh tick. And through the physical haze constant The corresponding one-dimensional point spread function filter kernel selected after gating is closer to the column drive execution surface in time and matches the scattering ability of the outermost anti-glare coating in space.

[0072] Specifically, step two can be used for engineering verification using three types of indicators: First, compare the actual mapped bounding box coordinates. First, check the continuity of the overlap with the reflected edge observed on the screen; second, check the angles in different directions of motion. Does the loaded one-dimensional point spread function filter kernel always satisfy the diagonal step error? Thirdly, when the vertical blanking parameter packet verification fails, display whether the driver correctly retains the previous frame result.

[0073] Step 3: Map the bounding box coordinates in reality Under the given premise of a one-dimensional point spread function filter kernel, the display driver outputs an asymmetric compensation drive to the panel pixel array, which has strong leading edge suppression, trailing relaxation release and chassis sudden change fuse protection.

[0074] Step two has already completed the coordinate extrapolation and loading of the one-dimensional point spread function filter kernel for object conjugation locally on the display driver. If step three still adopts the symmetrical method of uniformly applying pressure around the bounding box, although the secondary halo at the leading edge will be suppressed, a dark groove will form on the trailing side due to extrapolation jitter, link queuing, and pixel response differences. This dark groove does not originate from the external light field itself, but from the overly uniform compensation action, making the boundary break more easily noticeable to the human eye during scanning. Therefore, step three does not pursue uniform intensity suppression on all four sides, but rather focuses on the light spot movement vector. It actively breaks the symmetry in the direction of the trailing edge, giving priority to the limited driving capability for the leading edge, and leaving the trailing side to physical diffusion to complete the visual filling.

[0075] From the perspective of the execution entity, the display driver is responsible for mapping the actual bounding box coordinates. The one-dimensional point spread function filter kernel is converted into a pixel-level reverse suppression driving quantity. The source driver is responsible for superimposing this driving quantity onto the grayscale driving reference of the current row, and the gate driver is responsible for writing the superposition result into the panel array during the row scan timing.

[0076] For organic light-emitting diode (OLED) panel arrays, the reverse suppression drive quantity corresponds to the source data voltage reduction or the pulse width modulation duty cycle reduction; for active matrix liquid crystal panel arrays, the reverse suppression drive quantity corresponds to the source grayscale voltage bias. The main processor does not directly participate in pixel-by-pixel calculations, but continuously receives chassis accelerometer readings from the chassis domain communication bus in the background, and writes a high-priority interrupt flag to the vertical blanking area of ​​the next frame during sudden changes. With this configuration, step three retains both the timing advantages of the display driver's near-panel execution and the early warning of mechanical sudden changes.

[0077] The display driver first determines the actual mapped bounding box coordinates. The bounding box boundary and its outer extension band to be processed are determined. Then, the one-dimensional point spread function filter kernel loaded in step two is projected along the selected axis onto this outer extension band, forming a boundary inverse suppression base field. Subsequently, the display driver utilizes the spot motion vector... Distinguish between the leading edge and the trailing edge, and apply a trailing relaxation coefficient to the trailing edge. This maintains the leading edge at the first strength and reduces the trailing edge to the second strength. Finally, once the main processor detects that the chassis jerk has exceeded the fuse threshold, it writes a high-priority interrupt flag in the vertical blanking region; after the display driver reads this flag at the horizontal and vertical synchronization switching point, it immediately clears the usage result of the cross-domain communication lag time constant and simultaneously sets the trailing relaxation coefficient. Pulling back to 1, the asymmetric mode is converged into a symmetric protection mode. Thus, step three forms a single strand of "base field generation - directional splitting - mutation convergence".

[0078] Furthermore, the first step is to address how the one-dimensional point spread function filter kernel is mapped to the true bounding box coordinates. At the boundary. The display driver does not write the entire bounding box region as low brightness, but only forms a reverse suppression base field within the outer band of the bounding box boundary, because whitening distortion is mainly manifested as scattering up near the boundary, while the original interface content inside the bounding box still needs to remain readable. The display driver first determines the core tap length. Calculate the outer band width, then determine the normal distance from the boundary pixel to the nearest boundary on each scan line cut by the bounding box, and feed this normal distance to the kernel interpolation function. Specifically: Where: base field driving quantity : indicates that the display driver is in the _th ... The boundary position vector of the frame to be computed The generated reverse suppression base field driving quantity is within the range of reverse suppression driving quantities allowed by the panel; local reference driving quantity. : Represents the local reference reverse suppression drive quantity obtained by the display driver based on the current user interface drive code value, and takes a positive value; Nuclear tap length : Continuing with the definition of the length of the one-dimensional point spread function filter kernel from step two, the value is taken as an odd number greater than or equal to 3; summation index : Represents the discrete tap index of the one-dimensional point spread function filter kernel, with values ​​ranging from... arrive Integer; tap position index , indicating the first In the local template, the first A discrete tap; Filter tap coefficients : indicates the first The one-dimensional point spread function filter kernel of the frame is at the 1st The coefficients at each tap position, the sum of all filter tap coefficients is 1, and the value of each individual filter tap coefficient is non-negative; position vector : Indicates the pixel coordinates to be written on the current scan line, and the value is located in the display valid area; Normal distance : Represents a position vector Mapping to the actual bounding box coordinates The absolute value of the nearest boundary's sign distance, when located inside the boundary's outer band, is no greater than the outer band width; pixel pitch. Following the definition in step two, this represents the center distance between adjacent pixels, and the value is positive; mapping object The index of the local template or target area; kernel interpolation function : This function represents a first-order preserved interpolation of discrete taps based on the normal distance; its input is the difference between the normal distance and the discrete tap position in pixel increments. The normalized result falls within the closed interval. Press the button inside A linear change takes the value of zero if it falls outside the specified interval.

[0079] True mapping bounding box coordinates Indicates the first The bounding box coordinates after extrapolation and clipping to the displayable area in the current local refresh substep; its function is to serve as the normal distance function. A boundary reference object used to determine the current position vector. The distance relative to the current compensation region boundary; its value lies within the coordinate range of the display valid area. Interpolation independent variable. : Indicates normal distance With the Nominal position of each discrete tap The difference is based on pixel pitch. Normalized result; Frame index Indicates the current processing frame or the time sequence number under the current unified time base; Support domain constraints : Represents the current position vector Only when located within the effective support range of the current true mapping bounding box boundary extension band does it participate in the base field driving quantity. The generation of the kernel; its function is to limit the effective region of the one-dimensional point spread function filter kernel to the length of the kernel tap. and pixel pitch Within a jointly defined finite boundary band, avoid applying unnecessary reverse suppression drives to pixel locations far from the boundary.

[0080] In a preferred embodiment, the display driver first caches the bounding box boundary index of the current scan line and the previous scan line in the on-chip line buffer, and then calculates the base field driving amount by the fixed-point multiply-accumulate array in tap order. When an organic light-emitting diode (OLED) panel array uses a top-emitting structure, the base field driving quantity is directly superimposed on the source driving reference; for a bottom-emitting structure, the driving quantity is still superimposed in the same order, with a local reference driving quantity. This is derived from another set of panel process parameters. Therefore, the one-dimensional point spread function filter kernel is confined to the outer boundary band and a unified base field is reserved for subsequent directional splitting.

[0081] Furthermore, it addresses where the leading edge and trailing edge are located, and how the trailing edge is controlled for release. The display driver uses true-mapped bounding box coordinates. Using the center of the bounding box as a reference, and setting the current position vector... The displacement vector relative to the center of the bounding box and the light spot motion vector Make a projection.

[0082] When the projection amount When the projection is not less than zero, it is defined as the leading edge. A value less than zero is defined as a trailing edge. With this definition, the leading edge does not depend on the fixed left-right direction of the screen, but rather follows the movement vector of the light spot. They rotate together, so when the vehicle turns, pitches, and yaws simultaneously, the strength of the boundaries will not be misaligned.

[0083] Where: asymmetric driving quantity : Represents a position vector In the The final asymmetric reverse suppression driving quantity obtained in the frame takes values ​​between zero and the base field driving quantity. Between; base field driving quantity Using the aforementioned definition, this represents the reverse suppression base field driving force before leading-edge tailing separation; tailing relaxation coefficient. : Indicates the proportion of suppression retained on the trailing side, with values ​​within a closed interval. Its initial value is given by the joint calibration table of anti-glare coating haze level and panel response speed; position vector : Represents the coordinates of the current pixel in the screen coordinate system; center operator Using the definition from step one, this is used to map the true bounding box coordinates. Convert to bounding box center coordinates; True mapping bounding box coordinates Using the definition from step two, represents the current bounding box position; the light spot motion vector. Using the definitions from step one, represent the direction and speed of the light spot's motion; stability term : Represents a positive constant to avoid instability of the denominator near zero velocity, and takes a positive value less than the velocity dimension corresponding to one pixel pitch; Step gate function : Represents a sign-gated function; it takes the value 1 when the input is not less than zero and 0 when the input is less than zero. It is used to output a relaxation drive on the trailing side and maintain full-strength suppression on the leading edge side; Vector norm : Represents the motion vector of the light spot The modulus, taking non-negative values; Projection discriminant Indicates the first Frame position vector The normalized projection along the direction of light spot movement relative to the center of the actual mapped bounding box; its function is to distinguish whether the current position belongs to the leading edge or the trailing edge of the light spot movement; when When, it represents the position vector. Located on the leading edge side; when When, it represents the position vector. Located on the trailing side. Leading edge gating item. : Represents the discriminant for projection The leading edge selection result after step gating; its function is to output the full-strength base field driving quantity when the current position belongs to the leading edge side. Trailer door control item : Represents the discriminant for projection Taking the inverse number and then performing a step-gated tail selection result; its function is to output the tail relaxation coefficient when the current position belongs to the tail side. Decayed base field driving force. Time index. Represents the current processing frame or the time sequence number under the current unified time base; Step gate function This represents a sign-gated function; it takes the value 1 when the input is not less than zero and 0 when the input is less than zero; its function is to... Select full-intensity compression on the leading edge side. In this case, the tail side is relaxed and suppressed to achieve asymmetric drive. Directional allocation.

[0084] For example, when a vehicle turns right along an elevated ramp, and the reflection from the passenger-side window slides from the upper right to the left on the curved central control screen, the driver does not see a regular rectangular dark frame, but rather a compensation band with a harder front boundary and a softer rear boundary. The display driver does not change the one-dimensional point spread function filter kernel itself during this process; it only rewrites the driving quantity of the base field at different directional boundaries. The retention ratio. Tail relaxation coefficient. Take in closed interval The reason for this is that the trailing side still retains visible suppression within this range, but it does not completely cut off diffused light passing through the anti-glare coating. Thus, the direction of motion is truly embedded into the drive output chain.

[0085] This is further used to handle the risk of the extrapolated direction being instantly overturned in mechanically abrupt scenarios. The main processor continuously reads the output of the inertial measurement unit in the chassis domain communication bus, performs frame synchronization sampling on the gravity acceleration vector, and performs discrete derivatives on the continuous sampling points to obtain the chassis jerk. Once the chassis jerk exceeds the fuse threshold, the main processor sets the high-priority interrupt flag to one in the vertical blanking area of ​​the next frame. The display driver prioritizes reading this flag at the field synchronization boundary, no longer waiting for complete coordinate recalculation, but directly terminating the asymmetric mode.

[0086] Among these, the duration of mechanical mutations is shorter than the traditional image compensation replanning time. If the trailing motion continues along the original direction, it is most likely to result in an inverted edge. Therefore, the display driver connects the protection gating quantity in parallel to the extrapolation register write enable chain and the trailing relaxation coefficient selection chain, allowing mechanical mutations to directly interrupt the continuation of the historical direction. Specifically: Where: protection threshold quantity : indicates the first Whether the frame is allowed to continue with extrapolation and asymmetric release, with a value of 0 or 1; interrupt flag. : Indicates the high-priority interrupt flag written by the main processor to the vertical blanking region, with a value of 0 or 1. A value of 1 indicates that a mechanical sudden change has been detected; Step gate function Following the previous definition, the value is 1 when the input is non-negative and 0 when the input is negative. Circuit breaker threshold : Indicates the boundary value for triggering chassis jerk protection; the value is positive; chassis jerk. : indicates the first The degree of mechanical abrupt change corresponding to the frame is a non-negative value; where This represents the degravitational acceleration vector after attitude calculation to remove the gravity component; it also represents the hysteresis constant. : indicates that the display driver is in the _th ... The cross-domain communication lag time constant that the frame continues to use takes a value of 0 or the original cross-domain communication lag time constant. Cross-domain communication lag time constant Using the definition from step two, represents the time difference upon which the original extrapolation was based; the relaxation coefficient after protection. : Represents the tail relaxation coefficient after the protection action takes effect, and its value is located in the closed interval. Internal; tail relaxation coefficient Using the previous definition, this represents the compression ratio of the trailing side under normal conditions.

[0087] Gravity acceleration vector of the previous frame : indicates the first The degravity acceleration vector after attitude calculation to remove the gravity component from the frame; its function is to be compared with the current frame's degravity acceleration vector. Together, they constitute the change in acceleration between two adjacent frames, used to determine the chassis jerk. Current frame time interval Indicates the first Frame and the The unified time base interval between frames; its function is to convert the difference in gravity acceleration vector between two adjacent frames into chassis jerk. Its value is positive, preferably given by the unified timestamp of the inertial measurement unit, the system clock, or the time alignment result of the main processor. Vector norm operator The operator represents the Euclidean norm of the input vector; time index Indicates the current processing frame or the time sequence number under the current unified time base; In one implementation, the display driver displays the protection gate quantity. Two parallel hardware chains are connected: the first chain controls the write enable of the extrapolation result register, and the second chain controls the write enable of the protection gate. When it becomes zero, the new extrapolated offset is no longer written; the second chain controls the selection terminal of the tail relaxation coefficient register, when the protection gate quantity... When the value becomes zero, the register immediately outputs the relaxation coefficient after protection. In practice, after sudden braking or mechanical shocks caused by continuous potholes, the compensation boundary no longer shifts in its original direction, and the leading edge and trailing edge regain symmetrical enclosure. Therefore, mechanical shocks in the chassis domain are directly integrated into the display drive timing.

[0088] In use, the compensation drive output by the display driver is no longer based on the true mapped bounding box coordinates. Instead of equal pressure on all four sides, it uses the boundary reverse pressure base field as the base and the light spot motion vector as the base. For the splitting direction, use the tail relaxation coefficient Driven by asymmetric compensation for the release scale. The leading edge is stronger and the trailing edge is softer, thus blocking the secondary halo in front of the movement without generating a dark groove that is easily perceived by the human eye behind.

[0089] In engineering verification, high-speed imaging can be used to record the leading edge boundary stiffness, trailing edge continuity, and the timing of the compensating band returning to positive after a sudden change in chassis acceleration; the trailing relaxation coefficient can be turned off in the control path. Alternatively, disable the high-priority interrupt flag link and compare the locations where the trailing dark ditch and oblique afterimage appear.

[0090] Step 4: Distribute the asymmetric reverse suppression driving force. The original user interface drive signals are combined into panel execution pulses, and the light spot motion vector is corrected based on the in-vehicle light mixing feedback. Time decay weight constant Then, under static conditions, smoothly shut down the asynchronous extrapolation engine and restore the symmetrical boundary suppression state.

[0091] If the asymmetric reverse suppression driving quantity distribution output in step three is... If the compensation is only stored in the display driver's internal registers and not synthesized within the same line and field cycle as the original user interface drive signals, the compensation result cannot reach the physical photonic layer, and the aforementioned leading edge suppression and trailing relaxation will only remain at the algorithm layer. Conversely, if the compensation drive amount is simply superimposed directly onto the entire screen drive, the bounding box will no longer slide when stationary, and continuing to maintain asynchronous extrapolation and trailing relaxation will reduce local contrast. Therefore, step four organizes "voltage synthesis - closed-loop correction - state rollback" into a continuous single chain, so that dynamic sliding and stationary parking are both included in the same overall inventive concept.

[0092] The unified execution mechanism involves the coordinated efforts of the main processor, display driver, and physical panel array. The display driver is responsible for distributing the asymmetric inverse suppression drive in the current frame. The original user interface drive signals are time-series synthesized; the physical panel array is responsible for converting the synthesized pulse width modulation duty cycle or source-gate voltage pulse into emitted light; the main processor is responsible for correcting the time decay weight constant in the next frame's micro-parameter packet based on the in-vehicle light mixing results collected by the ambient light sensor array. The three execution entities are interconnected, and both the input and output objects have clear destinations. Therefore, this step is not abstract control, but an electro-optical closed loop directly coupled with the panel drive chain.

[0093] The display driver first reads the asymmetric reverse suppression drive quantity distribution given in step three. The original user interface drive signal of the current frame is mapped to a pulse width modulation duty cycle or source-gate voltage pulse according to the panel type; the physical panel array then performs electro-optic conversion and outputs the actual visible brightness inside the vehicle; the main processor then receives the mixed light equivalent brightness returned by the ambient light sensor array, compares it with the original user interface target brightness, forms a closed-loop contrast residual, and updates the time decay weight constant through a one-dimensional Kalman recursion; finally, the main processor continuously checks the light spot motion vector in the vertical blanking region. If the continuous static criterion is met, the shutdown flag is written into the subsequent parameter packet, and the display driver smoothly shuts down the asynchronous extrapolation engine accordingly, and the tail relaxation coefficient is adjusted. The value is increased to 1. Thus, step four completes the closed loop from the driving quantity to the photon and back to the parameter.

[0094] Furthermore, we first need to address how the display driver actually writes the results from step three into the panel. The display driver does not output an asymmetric reverse suppression drive quantity distribution independently. Instead, within each row of gate-opening windows, the original user interface drive signal and the asymmetric reverse suppression drive quantity are distributed. The data is fed into the same set of multiply-accumulate paths. For organic light-emitting diode (OLED) panel arrays, the multiply-accumulate result is first converted into a pulse-width modulation (PWM) duty cycle, and then mapped to the conduction time of the pixel driving transistor. For active matrix liquid crystal (LCD) panel arrays, the multiply-accumulate result is converted into the source grayscale voltage and written synchronously with the gate conduction signal. Although the final actuators differ between the two paths, the front-end synthesis object remains consistent, thus not disrupting the terminology defined in the preceding steps. Wherein: Path: Duty cycle or source data voltage; Path: Source grayscale voltage synchronized with gate write timing; Where: Composite duty cycle : indicates that the display driver is in the _th ... Frame-to-position vector The generated final normalized driving quantity takes values ​​within a closed interval. Original user interface drive signals : Represents the current position vector The normalized interface-driven quantity before compensation, with values ​​located in the closed interval. Panel coupling coefficient : Represents the distribution of asymmetric reverse suppression driving force The proportional coefficient for converting to normalized drive reduction is positive and is calibrated based on the panel electro-optical conversion curve. Asymmetric reverse suppression driving force distribution Using the definition from step three, let represent the current position vector. The final reverse suppression driving force at the point; saturation operator : This means cropping the input value to a closed interval. Operators within the range are set to 0 when less than 0 and 1 when greater than 1; In a preferred implementation, a two-stage latch chain is provided within the display driver, with the first stage latching the original user interface drive signal. Distribution of asymmetric reverse suppression driving quantity in the second-stage latch The fixed-point code. When line synchronization arrives, both are combined using the same multiply-accumulate array to generate a composite duty cycle. Then, the source drive latch and gate pulse generator are written sequentially according to the row address.

[0095] For example, when a vehicle enters the toll booth canopy under afternoon sidelight, the driver can directly see that the boundary of the reflective area is still suppressed, but the text and buttons on the toll interface are not darkened overall. This is because of the composite duty cycle. Only in the asymmetric reverse suppression driving quantity distribution The original user interface drive signals are subtracted from the covered local area. The boundary compensation in step three is rewritten into a real driver that the panel can execute, rather than uniformly reducing the brightness of the entire screen.

[0096] Furthermore, after the light emitted by the physical panel array is superimposed with the external ambient light inside the vehicle, the driver actually perceives the mixed light result. Therefore, the main processor does not directly use the target brightness of the panel for correction, but instead obtains the equivalent brightness of the mixed light inside the vehicle through the ambient light sensor array. The ambient light sensor array is preferably arranged on the upper bezel of the curved screen, the inner edge of the instrument panel, and near the air duct of the center console. Multiple measuring points are first converted into a unified equivalent brightness of the mixed light by installing a weight matrix.

[0097] The main processor feeds the comparison result into a one-dimensional Kalman recursion. The recursive state is not the pixel value, but the time decay weight constant for writing the micro-parameter packet to the next frame. This indirectly corrects the light spot motion vector. The reference gain in subsequent extrapolation.

[0098] Where: Time decay weight constant : Represents the updated weights generated by the main processor for the next frame's micro-parameter packet, with values ​​located in a closed interval. Time decay weighting constant : Represents the prior weight estimate currently in use for the current frame, with values ​​located within the closed interval. Kalman gain : Represents the current correction gain obtained by the main processor through one-dimensional Kalman recursion, with values ​​located within a closed interval. ;in and Let these represent the prior and posterior covariances, respectively. Represents the process noise covariance. Represents the observation noise covariance; Equivalent brightness of mixed light : Represents the actual mixed light brightness inside the vehicle after the ambient light sensor array is converted by the installation weight matrix, and the value is positive; where This indicates the number of sensors involved in the synthesis. Indicates the first The installation weight of each sensor, Indicates the first The sensor at the first Measured brightness of the frame; brightness of the target interface : Represents the target brightness planned for the original user interface output in the current frame, and takes a positive value; stability term : Represents a positive constant to prevent instability when the denominator approaches zero; its value is less than the target interface brightness. The lower bound; the minimum weight lower bound : Represents the minimum value to which the time decay weight constant is allowed to decrease, and its value lies within the open interval. ; Saturation operator : indicates that the input is clipped to a closed interval. Operators within the loop. Closed-loop observation residuals. Indicates the first Frame interior light mixing equivalent brightness With the target interface brightness The normalized bias between them; its function is as an observation in the one-dimensional Kalman recursion, used to correct the prior estimate of the time decay weight constant; its value can be positive or negative, when Higher than When to take the right time Below Take the negative value. Posterior weight estimation from the previous frame. : indicates the first The posterior estimate of the time decay weight constant obtained after fusing the observation residuals of the frame; its role is as the first... Frame prior weight estimation The starting point of the recursion; its value lies within a closed interval. .

[0099] Posterior covariance of the previous frame : indicates the first The posterior estimated covariance of the frame obtained after fusing the observation residuals; its role is as the first... Frame prior covariance The recursive basis, and the process noise covariance Together they determine the prior uncertainty level of the current frame; Summation index This indicates the sensor number index involved in the synthesis of equivalent brightness in the vehicle interior; the value ranges from 1 to... Integer. Time index. Indicates the current processing frame or the time sequence number under the current unified time base; In practice, the main processor uses the time decay weight constant. As a one-dimensional state variable, the closed-loop contrast residual As an observation measure, the quantization noise level of the ambient light sensor array is used as the Kalman gain input for observation confidence. The generation chain. The initial value of the time decay weight constant is given by the calibration table, and the initial covariance is given by the factory conformance test; the observation noise covariance... The process noise covariance is determined by the current confidence level of the ambient light sensor. The weights are determined by the fluctuations in weight updates over several preceding frames. After this processing, the main processor sends the display driver not a new full-screen image, but only the updated time-decay weight constant. It is still related to the light spot motion vector Both are in the same micro parameter package.

[0100] For example, when a vehicle moves from a brightly lit elevated road section into a tree-lined road section, the driver will see a smooth transition in overall screen readability, rather than an abrupt jump in the compensation band. This is because of the time decay weight constant. It is slowly corrected under the effect of light mixing feedback.

[0101] When in use, the parameter domain compensation formed in steps one to three does not revert to the pixel domain in the closed-loop stage, thus maintaining the advantage of lightweight transmission.

[0102] Furthermore, if the vehicle is parked waiting at a red light or remains stationary for a short period due to external reflections, continuing to operate the asynchronous push engine will cause the display driver to unnecessarily advance the surround box along the historical direction, while the trailing relaxation coefficient... The asymmetric release is still retained, which will cause the boundary compensation to deviate from the static high contrast requirement.

[0103] To this end, the main processor continuously monitors the motion vector of the light spot in the vertical blanking region. After a number of consecutive frames meet the stillness threshold, a shutdown flag is written to the display driver.

[0104] After the display driver receives the shutdown flag, it first freezes the cross-domain communication hysteresis constant. Extrapolation offset, and then the tail relaxation coefficient. The process smoothly pushes the material towards 1 through a first-order lifting process, thus returning it to the state of symmetric boundary suppression.

[0105] Where: static determination quantity : indicates the first Whether a frame satisfies the condition of continuous stillness, with a value of 0 or 1; still frame window : Represents the number of consecutive frames involved in the stillness detection, with a value greater than or equal to 3; cumulative index. : Indicates the still frame window The historical frame offset involved in continuous determination has a value ranging from 0 to... Integers; gated functions : Represents a step gate function, taking the value 1 when the input is not less than zero and 0 when the input is less than zero; Rest threshold : Represents the static determination boundary of the light spot motion vector magnitude, and takes a positive value; light spot motion vector Using the definition from step one, it represents the motion direction and motion rate of the current frame and the previous few frames; Vector norm : Represents the motion vector of the light spot The modulus length, taking non-negative values; the tail relaxation coefficient. : Represents the trailing relaxation coefficient for the next frame, with values ​​within a closed interval. Tail relaxation coefficient Using the definition from step three, this represents the reverse suppression retention ratio on the trailing side of the current frame, with a value within the closed interval. ; Step-up coefficient : Indicates the tail relaxation coefficient The smooth speed that pushes towards 1, with values ​​located in the open interval. ; In a preferred embodiment, the main processor determines the stationary state. This is written to the next frame's micro-parameter packet. After the display driver reads this value during vertical blanking, it first disables the write enable of the asynchronous extrapolation engine, and then updates the trailing relaxation coefficient. For example, when a vehicle is stopped at an intersection and the reflection from the side window is fixed in a corner of the curved central control screen, the driver will see the previously directional compensation band gradually transform into a balanced, suppressed boundary around the perimeter. This change is smooth over several frames without any sudden edge flips. For assemblies using organic light-emitting diode (OLED) panel arrays, this back-off process directly changes the pulse width modulation duty cycle; for assemblies using active matrix liquid crystal (LCD) panel arrays, this back-off process changes the source grayscale voltage and the gate write window, but the state determination logic and terminology remain unchanged. Specifically, the system can maintain directional compensation during dynamic sliding and restore high-contrast symmetrical display when stationary.

[0106] Specifically, the display driver distributes the asymmetric reverse suppression drive quantity. The original user interface drive signals are synthesized into a truly executable panel driver within the same line and field beat. The physical panel array then transcribes this driver into visible light output, and the main processor finally converges the light mixing feedback into a time-decay weighted constant. The update amount, and through the static judgment amount Trigger state rollback.

[0107] Furthermore, the composite duty cycle can be... Linear continuity, mixed light equivalent brightness With the target interface brightness The direction of residual change and the static judgment quantity Post-trigger trail relaxation coefficient The pullback process is used as an evaluation indicator; the time decay weight constant can be turned off in the comparison path. The system includes feedback updates and static condition rollback functions to observe the location of compensation lag and static boundary skew.

[0108] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An adaptive brightness compensation driving method for a low-glare display component, characterized in that: include, The main processor calculates the bounding box reference coordinates and the spot motion vector based on the ambient light field vector output by the vehicle-mounted light sensor array, the pose data of the inertial measurement unit, and the three-dimensional topological normal matrix of the curved screen, and encodes them into a micro parameter packet embedded in the vertical blanking area; the display driver parses the micro parameter packet, deduces the real mapping bounding box coordinates based on the spot motion vector and the cross-domain communication lag time constant, and loads a one-dimensional point spread function filter kernel based on the physical haze constant of the anti-glare coating; The display driver generates an asymmetric inverse suppression drive quantity based on the real mapped bounding box coordinates, the one-dimensional point spread function filter kernel, and the direction of the spot motion vector; the display driver then combines the asymmetric inverse suppression drive quantity with the UI drive signal to drive the physical panel array.

2. The adaptive brightness compensation driving method according to claim 1, characterized in that: The main processor determines the ambient light field vector based on the sampling value, installation direction, and sensitivity calibration coefficient of the vehicle-mounted optical sensor array, and calculates the bounding box reference coordinates by combining the three-dimensional topological normal matrix of the curved screen. When the inertial measurement unit data is abnormal, the cockpit DMS infrared camera is called to extract the displacement of the driver's corneal reflected light spot and map it into a light spot motion vector. Data abnormalities include data packet loss and timestamp confidence below the safety threshold.

3. The adaptive brightness compensation driving method according to claim 2, characterized in that: The micro parameter packet includes frame index, bounding box reference coordinates, spot motion vector, time decay weight constant, degradation flag, high priority interrupt flag, and cyclic redundancy check field. The main processor writes the micro parameter packet into the vertical blanking area and sends it to the display driver through the underlying video transmission link, where the display driver latches and parses it during the local refresh cycle.

4. The adaptive brightness compensation driving method according to claim 3, characterized in that: In the blind-end coordinate system, the display driver takes the bounding box reference coordinates as the starting point and performs forward extrapolation based on the spot motion vector, time decay weight constant, and cross-domain communication lag time constant. It also includes the write packet delay, parsing delay, line scan delay, and panel electro-optic response delay in the cross-domain communication lag time constant to obtain the true mapped bounding box coordinates.

5. The adaptive brightness compensation driving method according to claim 4, characterized in that: The display driver reads the physical haze constant of the anti-glare coating from the read-only configuration area and, based on the physical haze constant and the direction of the spot motion vector; Select the corresponding one-dimensional point spread function filter kernel in the internal static cache, and complete the loading only when the two-dimensional diagonal step error of the selected filter kernel is not greater than the optical homogenization tolerance calculated from the physical haze constant.

6. The adaptive brightness compensation driving method according to claim 5, characterized in that: The display driver forms an outer band based on the boundary of the real-mapped bounding box coordinates, and projects a one-dimensional point spread function filter kernel along a selected axis onto the outer band. It generates a boundary reverse suppression drive based on the normal distance from the pixel position to the nearest boundary, where the boundary reverse suppression drive only acts on the panel pixel array corresponding to the outer band.

7. The adaptive brightness compensation driving method according to claim 6, characterized in that: The display driver uses the center of the true mapped bounding box coordinates and the direction of the spot motion vector as the basis for determining the boundary orientation. The boundary where the projection result is located on the forward side is defined as the leading edge side, and the boundary where the projection result is located on the backward side is defined as the trailing edge side. The driving amount is suppressed by the boundary with the first intensity on the leading edge side and attenuated to the second intensity on the trailing edge side according to the trailing relaxation coefficient.

8. The adaptive brightness compensation driving method according to claim 7, characterized in that: The main processor performs frame synchronization differential on the gravity acceleration vector output by the inertial measurement unit to obtain the chassis jerk. When the chassis jerk reaches the fuse threshold, a high-priority interrupt flag is written in the next vertical blanking region. After the display driver reads the high-priority interrupt flag, it stops the extrapolation update of the real mapped bounding box coordinates and resets the trailing relaxation coefficient to one.

9. The adaptive brightness compensation driving method according to claim 8, characterized in that: The display driver combines the asymmetric reverse suppression drive quantity and the UI drive signal in the horizontal and vertical synchronization timing to generate a pulse width modulation duty cycle and form a corresponding source gate voltage pulse, which is then written to the physical panel array. The main processor reads the brightness of multiple measurement points of the ambient light sensor array, calculates the equivalent brightness of the mixed light inside the vehicle, and compares it with the brightness of the target interface to obtain the closed-loop contrast residual.

10. The adaptive brightness compensation driving method according to claim 9, characterized in that: The main processor uses the closed-loop contrast residual as an observation, performs Kalman filtering recursion on the time decay weight constant, and writes the updated time decay weight constant into the subsequent micro parameter package; when the magnitude of the light spot motion vector of multiple consecutive frames is lower than the static threshold, the display driver turns off asynchronous extrapolation and smoothly increases the trailing relaxation coefficient to one.

Images