A screen area adaptive geometric correction method and system
By using user-defined initial reference points and interpolation to generate dense auxiliary points, a two-dimensional mesh is constructed and coordinate mapping is performed. This solves the problem of complex deformation in the screen area captured by the camera, achieving high-precision adaptive geometric correction of the screen area. It has wide applicability and is computationally lightweight.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TIANJIN HUAJUWULIAN TECHNOLOGY LTD
- Filing Date
- 2026-04-08
- Publication Date
- 2026-06-19
AI Technical Summary
Existing technologies struggle to accurately reconstruct standard rectangular images from screen areas captured by cameras when complex geometric deformations exist, impacting the accuracy of lighting effect synchronization and immersive experience. Furthermore, relying on pre-trained models presents generalization problems and computational overhead.
By using user-defined initial reference points, dense auxiliary points are generated using boundary interpolation and linear interpolation methods to construct a two-dimensional grid. Irregular screen areas are then transformed into standard rectangles through coordinate mapping relationships to generate an adaptive geometrically corrected image.
It achieves high-precision adaptive geometric correction of screen area, reduces the requirements for device computing power, has wide applicability, is computationally lightweight, meets the real-time processing requirements of high frame rate video streams, and avoids model generalization problems.
Smart Images
Figure CN122243832A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image communication technology, and in particular to a screen area adaptive geometric correction method and system. Background Technology
[0002] With the development of immersive entertainment and smart homes, ambient lighting synchronization technology based on screen content analysis (such as game linkage and ambient lighting effects) is becoming increasingly popular. The core of this type of application lies in: capturing the display screen in real time through image acquisition devices (such as cameras), accurately extracting the color information of specific areas from it, and mapping it to the surrounding lighting effects.
[0003] However, in actual deployments, cameras are often difficult to install directly facing the center of the screen (e.g., placed below or to the side of the TV), resulting in severe perspective distortion and geometric distortion in the captured screen images. What is originally a rectangular screen area appears as an irregular quadrilateral or even a curved polygon in the image. Directly performing color analysis on this area will introduce a large number of errors, seriously affecting the accuracy and immersiveness of lighting effect synchronization.
[0004] Therefore, the key technical prerequisite for ensuring the accuracy of subsequent color extraction is how to reconstruct the screen area with complex geometric deformation captured by the camera into a standard rectangular image in real time and with high precision.
[0005] Existing technologies use the following methods to reconstruct standard rectangular images in real time and with high precision from screen areas containing complex geometric deformations captured by cameras: Image segmentation and boundary detection: The input image is segmented using a pre-trained image segmentation model (such as the YOLO series), the "fisheye image" region is identified, and the boundary points are detected. Boundary point determination and region subdivision: The detected boundaries are linearly completed and denoised to form a boundary framework. According to predefined rules (such as finding feature points such as corner points and midpoints of long sides), a series of "pre-defined points" are determined on the boundary. Then, based on the mapping relationship between these pre-defined points, more internal points are interpolated within the display area, subdividing the entire screen area into multiple quadrilateral grids.
[0006] Partition correction: For each subdivided quadrilateral sub-region, the perspective transformation matrix is calculated separately, and it is corrected into a regular rectangle. Finally, the complete corrected image is synthesized.
[0007] The advantage of this technology lies in its ability to improve the correction of complex distortions through region subdivision. However, its core relies on a well-trained image segmentation model to initially obtain the screen region, which introduces dependence on the training dataset, model generalization problems on different devices and environments, and certain computational overhead. In addition, its boundary point fixing rules are relatively fixed, and the fitting accuracy and flexibility may be limited for extremely irregular non-convex screen borders. Summary of the Invention
[0008] The technical problem to be solved by this invention is to provide a screen region adaptive geometric correction method and system. By using a small number of user-calibrated points, a dense and smooth set of points describing the screen edge is generated using a specific interpolation algorithm, thereby fitting screen regions of arbitrary irregular shapes with high precision. Then, by constructing a mapping relationship from source space to target space, the irregular region is transformed into a standard rectangular image, and finally a standardized screen region adaptive geometric correction image that corrects perspective and edge distortion is generated.
[0009] This invention is achieved through the following technical solution: A screen area adaptive geometric correction method includes the following steps: S1: The system guides the user to select the four vertices of the effective display area of the source space screen image captured by the camera and the midpoint of the top border as the initial reference points for describing the screen distortion contour. S2: Using the initial reference point as the initial boundary control point, multiple auxiliary points for the upper edge of the screen are generated by boundary interpolation, resulting in an irregular upper edge curve and a sequence of upper edge points. Multiple auxiliary points for the lower edge of the screen are generated by linear interpolation, resulting in an irregular lower edge curve and a sequence of lower edge points. S3: Connect the corresponding points of the top and bottom edge point sequences of the screen to form multiple vertical connection lines. Then, apply an interpolation algorithm to each vertical connection line to generate multiple vertical connection line auxiliary points. Connect the corresponding vertical connection line auxiliary points on different vertical connection lines to form multiple horizontal connection lines. Finally, generate a two-dimensional source grid consisting of an irregular top edge curve, an irregular bottom edge curve, multiple vertical connection lines, and multiple horizontal connection lines. S4: Using a standard rectangular region corresponding to the screen area in the source space screen image as the transformation target image of the target space, the two-dimensional source grid and the transformation target image of the target space are mapped and aligned to obtain the coordinate mapping relationship from the source space to the target space. S5: Resample the source space screen image captured in real time by the camera. Based on the coordinate mapping relationship from the source space to the target space, query the coordinates of each pixel in the transformed target image in the source space screen image, and finally generate an adaptive geometric correction image for the screen area.
[0010] Furthermore, in step S2, multiple auxiliary points for the upper edge of the screen are generated using boundary interpolation according to equation (1): (1); in: This indicates the first [unclear] generated by screen top border interpolation. Coordinates of auxiliary points This indicates the coordinates of the starting point of the current segment at the top edge of the screen. This indicates the coordinates of the midpoint of the top edge of the screen. This indicates the number of auxiliary points on the top edge of the screen.
[0011] Furthermore, in step S2, multiple auxiliary points for the lower border of the screen are generated using linear interpolation according to equation (2): (2); in: This indicates the first value generated by interpolation of the bottom border of the screen. Coordinates of auxiliary points This indicates the coordinates of the starting point of the bottom border of the screen. This indicates the coordinates of the end point of the bottom border of the screen. This indicates the number of auxiliary points on the bottom border of the screen.
[0012] Furthermore, in step S3, an interpolation algorithm is applied to each longitudinal connecting line to generate multiple auxiliary points according to equation (3): (3); in: Indicates the first The first column on the vertical connecting line Line auxiliary point coordinates, This indicates the first [unclear] generated by screen top border interpolation. Coordinates of auxiliary points Indicates the number of vertical grid lines. This indicates the first value generated by interpolation of the bottom border of the screen. Coordinates of auxiliary points.
[0013] Furthermore, in step S4, the method for mapping and aligning the two-dimensional source mesh with the transformed target image to obtain the coordinate mapping relationship from the source space to the target space is as follows: For each horizontal line segment in the pixel grid of the target image, the corresponding coordinates of each pixel in the source space are calculated using the bilinear transformation algorithm according to equation (4), forming a pixel mapping relationship from the source space to the target space: (4); in: This represents the horizontal line segment in each pixel grid of the target image being transformed. The corresponding coordinates of each pixel in the source space This represents the starting pixel of the horizontal line segment in each pixel grid of the target image being transformed. Indicates direction symbol, This represents the floor function. This represents the change in source space coordinates corresponding to one unit distance in the target space.
[0014] A screen region adaptive geometric correction system is used to perform a screen region adaptive geometric correction method as described in any one of the above, characterized in that it includes an initial reference point acquisition module, an irregular outer border construction module, a two-dimensional source mesh construction module, a coordinate mapping relationship construction module, and a screen region adaptive geometric correction image generation module. The initial reference point acquisition module is used to select the four vertices of the effective display area of the source space screen image and the midpoint of the upper border as the initial reference points for describing the screen distortion contour. The irregular outer border construction module is used to generate multiple screen upper border auxiliary points using the initial reference point as the initial boundary control point and the boundary interpolation method to obtain an irregular screen upper border curve and a screen upper border point sequence. It also uses the linear interpolation method to generate multiple screen lower border auxiliary points to obtain an irregular screen lower border curve and a screen lower border point sequence. The two-dimensional source mesh construction module is used to connect the points at corresponding positions of the irregular screen top edge point sequence and the irregular screen bottom edge point sequence in pairs to form multiple vertical connection lines. Then, an interpolation algorithm is applied to each vertical connection line to generate multiple vertical connection line auxiliary points. The vertical connection line auxiliary points corresponding to different vertical connection lines are connected to form multiple horizontal connection lines. Finally, a two-dimensional source mesh is generated, which consists of an irregular screen top edge curve, an irregular screen bottom edge curve, multiple vertical connection lines, and multiple horizontal connection lines. The coordinate mapping relationship construction module is used to take a standard rectangular area corresponding to the screen area in the source space screen image as the transformation target image of the target space, and map and align the two-dimensional source grid with the transformation target image to obtain the coordinate mapping relationship from the source space to the target space. The screen region adaptive geometric correction image generation module is used to resample the source space screen image captured in real time by the camera. Based on the coordinate mapping relationship from the source space to the target space, it queries the coordinates of each pixel in the transformed target image in the source space screen image, and finally generates the screen region adaptive geometric correction image.
[0015] Beneficial effects of the invention: The screen area adaptive geometric correction method and system provided by this invention have the following advantages: 1. This invention completely eliminates the reliance on pre-trained deep learning models, requiring only a simple user calibration. It has wider applicability, is easier to deploy, reduces the requirements for device computing power, avoids model generalization problems, and enables the solution to be installed and used immediately on various hardware platforms and in unfamiliar environments.
[0016] 2. This invention actively generates dense auxiliary points as contour points between initial calibration points through an algorithm, which can more smoothly and accurately fit irregular shapes such as curved screens and borders with complex chamfers. Compared with existing technologies that mainly select discrete feature points on the detected boundaries, this invention improves accuracy from the data source and achieves higher fitting accuracy for irregular shapes such as curved screens and borders with complex chamfers.
[0017] 3. It avoids the forward inference computation of running an image segmentation model. The overall algorithm mainly consists of interpolation and coordinate transformation, which is computationally lightweight and has a small overhead, making it better able to meet the real-time processing requirements of high frame rate video streams. Attached Figure Description
[0018] Figure 1 This is a schematic diagram of the process of this invention. Detailed Implementation
[0019] A flowchart of a screen area adaptive geometric correction method is shown below. Figure 1 As shown, it includes the following steps: S1: The system guides the user to select the four vertices of the effective display area of the source space screen image captured by the camera and the midpoint of the top border as the initial reference points for describing the screen distortion contour. Specifically, the system guides users through a human-computer interaction interface to select five key points within the effective display area of the screen in the source space image captured by the camera. These five key points include the top-left corner, the midpoint of the top bezel, the top-right corner, the bottom-left corner, and the bottom-right corner. The calibration of the midpoint of the top bezel is particularly important because, in actual installations, the camera is usually located above the screen, causing the captured top bezel to exhibit an arc-shaped distortion. Two vertices alone cannot accurately describe this arc-shaped contour. These five key points serve as the initial reference points for describing the screen distortion contour and are the foundation for all subsequent calculations.
[0020] This step simplifies the complex problem of irregular edge detection into a small amount of deterministic user input, which is a prerequisite for achieving lightweight correction.
[0021] The source space mentioned later refers to the space where the original image captured by the camera is located, and its screen area has geometric distortion.
[0022] S2: Using the initial reference point as the initial boundary control point, multiple auxiliary points for the upper edge of the screen are generated by boundary interpolation, resulting in an irregular upper edge curve and a sequence of upper edge points. Multiple auxiliary points for the lower edge of the screen are generated by linear interpolation, resulting in an irregular lower edge curve and a sequence of lower edge points. Since the camera is usually located above the screen during actual installation, the top edge of the captured image will appear as an arc-shaped distortion. Therefore, the top edge can be divided into a left half and a right half and processed in segments. The left half is from the top left corner to the midpoint of the top edge, and the right half is from the midpoint of the top edge to the top right corner.
[0023] Specifically, multiple auxiliary points for the top edge of the screen can be generated using boundary interpolation according to equation (1): (1); in: This indicates the first [unclear] generated by screen top border interpolation. Coordinates of auxiliary points This indicates the coordinates of the starting point of the current segment at the top edge of the screen. This indicates the coordinates of the midpoint of the top edge of the screen. This indicates the number of auxiliary points on the top edge of the screen.
[0024] This boundary interpolation method uses quadratic curve interpolation to generate multiple auxiliary points on the top edge of the screen, which can smoothly fit the irregular arc-shaped top edge.
[0025] As for the bottom border, that is, the area between the bottom left and bottom right corners, since the distortion is relatively small, a linear interpolation method can be used to generate multiple auxiliary points for the bottom border of the screen between the two points.
[0026] Specifically, multiple auxiliary points for the bottom border of the screen can be generated according to equation (2): (2); in: This indicates the first value generated by interpolation of the bottom border of the screen. Coordinates of auxiliary points This indicates the coordinates of the starting point of the bottom border of the screen. This indicates the coordinates of the end point of the bottom border of the screen. This indicates the number of auxiliary points on the bottom border of the screen.
[0027] By using the linear interpolation method described above, multiple auxiliary points for the bottom border of the screen are generated between two points, which can smoothly fit the bottom border.
[0028] Once the fitting is complete, the system can obtain the upper and lower border point sequences of the screen.
[0029] S3: Connect the points at corresponding positions of the irregular upper and lower border points of the screen in pairs to form multiple vertical connection lines. Then, apply an interpolation algorithm to each vertical connection line to generate multiple vertical connection line auxiliary points. Connect the corresponding vertical connection line auxiliary points on different vertical connection lines to form multiple horizontal connection lines. Finally, generate a two-dimensional source grid composed of an irregular upper and lower border curve, an irregular lower border curve, multiple vertical connection lines, and multiple horizontal connection lines. Specifically, multiple auxiliary points for the longitudinal connecting lines can be generated according to equation (3): (3); in: Indicates the first The first column on the vertical connecting line Line auxiliary point coordinates, This indicates the first [unclear] generated by screen top border interpolation. Coordinates of auxiliary points This indicates the number of vertical grid lines, including vertical connecting lines and the top and bottom bezel curves of the screen. This indicates the first value generated by interpolation of the bottom border of the screen. Coordinates of auxiliary points.
[0030] Through the aforementioned horizontal and vertical interpolation, the system ultimately generates a two-dimensional grid, denoted as the source grid. Each node of this grid precisely corresponds to a specific location on the screen in the original distorted image, completely describing the irregular shape of the screen region in the source space.
[0031] S4: Using a standard rectangular region corresponding to the screen area in the source space screen image as the target image of the target space, the two-dimensional source grid and the target image of the target space are mapped and aligned to obtain the coordinate mapping relationship from the source space to the target space. The system can predefine a standard rectangular region corresponding to the screen area in the source space screen image, and transform the target image in the target space. The standard rectangular region is within the target space.
[0032] The target space refers to the space where the standard rectangular image output after correction is located. Within this space, the screen area has a regular shape and no distortion.
[0033] Specifically, the following method can be used to map and align the two-dimensional source mesh with the transformed target image to obtain the coordinate mapping relationship from the source space to the target space: For each horizontal line segment in the pixel grid of the target image, the corresponding coordinates of each pixel in the source space are calculated using the bilinear transformation algorithm according to equation (4), forming a pixel mapping relationship from the source space to the target space: (4); in: This represents the horizontal line segment in each pixel grid of the target image being transformed. The corresponding coordinates of each pixel in the source space This represents the starting pixel of the horizontal line segment in each pixel grid of the target image being transformed. Indicates direction symbols, used to handle coordinate incrementing or decrementing mappings. This represents the floor function. This represents the change in source space coordinates corresponding to one unit distance in the target space; it is a scaling factor. , This represents the number of pixels in the horizontal line segments of each pixel grid in the target image being transformed. This represents the endpoint pixel of the horizontal line segment in each pixel grid of the target image being transformed. This represents the starting pixel of the horizontal line segment in each pixel grid of the target image being transformed.
[0034] The algorithm described above can pre-calculate the coordinates of all target space pixels in the source space, forming a coordinate mapping relationship. This mapping relationship defines the complete geometric correspondence between the distorted image in the source space and the corrected image in the target space, thus laying the foundation for subsequent image transformation and output.
[0035] S5: Resample the source space screen image captured in real time by the camera. Based on the coordinate mapping relationship from the source space to the target space, query the coordinates of each pixel in the transformed target image in the source space screen image, and finally generate an adaptive geometric correction image for the screen area.
[0036] This invention, through the above method, requires only a simple user calibration, making it more applicable and easier to deploy. It reduces the requirements for device computing power, completely eliminates the dependence on pre-trained deep learning models, avoids model generalization problems, and avoids forward inference calculations of image segmentation models. The overall algorithm mainly consists of interpolation and coordinate transformation, making it computationally lightweight and less expensive. It can better meet the real-time processing needs of high frame rate video streams, enabling the solution to be installed and used immediately on various hardware platforms and in unfamiliar environments.
[0037] This invention uses an algorithm to actively generate dense auxiliary points as contour points between initial calibration points, which can more smoothly and accurately fit irregular shapes such as curved screens and borders with complex chamfers. It improves accuracy from the data source and has higher fitting accuracy for irregular shapes such as curved screens and borders with complex chamfers. Finally, the screen area adaptive geometric correction image is a standardized rectangular image that corrects perspective and edge distortion.
[0038] A screen region adaptive geometric correction system is used to perform a screen region adaptive geometric correction method as described in any one of the above, characterized in that it includes an initial reference point acquisition module, an irregular outer border construction module, a two-dimensional source mesh construction module, a coordinate mapping relationship construction module, and a screen region adaptive geometric correction image generation module. The initial reference point acquisition module is used to select the four vertices of the effective display area of the source space screen image and the midpoint of the upper border as the initial reference points for describing the screen distortion contour. The irregular outer border construction module is used to generate multiple screen upper border auxiliary points using the initial reference point as the initial boundary control point and the boundary interpolation method to obtain an irregular screen upper border curve and a screen upper border point sequence. It also uses the linear interpolation method to generate multiple screen lower border auxiliary points to obtain an irregular screen lower border curve and a screen lower border point sequence. The two-dimensional source mesh construction module is used to connect the points at corresponding positions of the irregular screen top edge point sequence and the irregular screen bottom edge point sequence in pairs to form multiple vertical connection lines. Then, an interpolation algorithm is applied to each vertical connection line to generate multiple vertical connection line auxiliary points. The vertical connection line auxiliary points corresponding to different vertical connection lines are connected to form multiple horizontal connection lines. Finally, a two-dimensional source mesh is generated, which consists of an irregular screen top edge curve, an irregular screen bottom edge curve, multiple vertical connection lines, and multiple horizontal connection lines. The coordinate mapping relationship construction module is used to take a standard rectangular area corresponding to the screen area in the source space screen image as the transformation target image of the target space, and map and align the two-dimensional source grid with the transformation target image to obtain the coordinate mapping relationship from the source space to the target space. The screen region adaptive geometric correction image generation module is used to resample the source space screen image captured in real time by the camera. Based on the coordinate mapping relationship from the source space to the target space, it queries the coordinates of each pixel in the transformed target image in the source space screen image, and finally generates the screen region adaptive geometric correction image.
[0039] In summary, the screen region adaptive geometric correction method and system provided by this invention can perform high-precision fitting of irregular screen borders, and finally generate a standardized screen region adaptive geometric correction image that corrects perspective and edge distortion. Moreover, it does not rely on AI models, has wider applicability, is easier to deploy, is more computationally lightweight, and has stronger real-time performance.
[0040] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A screen area adaptive geometric correction method, characterized in that: Includes the following steps: S1: The system guides the user to select the four vertices of the effective display area of the source space screen image captured by the camera and the midpoint of the top border as the initial reference points for describing the screen distortion contour. S2: Using the initial reference point as the initial boundary control point, multiple auxiliary points for the upper edge of the screen are generated by boundary interpolation, resulting in an irregular upper edge curve and a sequence of upper edge points. Multiple auxiliary points for the lower edge of the screen are generated by linear interpolation, resulting in an irregular lower edge curve and a sequence of lower edge points. S3: Connect the corresponding points of the top and bottom edge point sequences of the screen to form multiple vertical connection lines. Then, apply an interpolation algorithm to each vertical connection line to generate multiple vertical connection line auxiliary points. Connect the corresponding vertical connection line auxiliary points on different vertical connection lines to form multiple horizontal connection lines. Finally, generate a two-dimensional source grid consisting of an irregular top edge curve, an irregular bottom edge curve, multiple vertical connection lines, and multiple horizontal connection lines. S4: Using a standard rectangular region corresponding to the screen area in the source space screen image as the target image of the target space, the two-dimensional source grid and the target image of the target space are mapped and aligned to obtain the coordinate mapping relationship from the source space to the target space. S5: Resample the source space screen image captured in real time by the camera. Based on the coordinate mapping relationship from the source space to the target space, query the coordinates of each pixel in the transformed target image in the source space screen image, and finally generate an adaptive geometric correction image for the screen area.
2. The screen area adaptive geometric correction method according to claim 1, characterized in that: In step S2, multiple auxiliary points for the upper edge of the screen are generated using boundary interpolation according to equation (1): （1）； in: This indicates the first [unclear] generated by screen top border interpolation. Coordinates of auxiliary points This indicates the coordinates of the starting point of the current segment at the top edge of the screen. This indicates the coordinates of the midpoint of the top edge of the screen. This indicates the number of auxiliary points on the top edge of the screen.
3. The screen area adaptive geometric correction method according to claim 1, characterized in that: In step S2, multiple auxiliary points for the bottom border of the screen are generated using linear interpolation according to equation (2): （2）； in: This indicates the first value generated by interpolation of the bottom border of the screen. Coordinates of auxiliary points This indicates the coordinates of the starting point of the bottom border of the screen. This indicates the coordinates of the end point of the bottom border of the screen. This indicates the number of auxiliary points on the bottom border of the screen.
4. The screen area adaptive geometric correction method according to claim 1, characterized in that: In step S3, an interpolation algorithm is applied to each longitudinal connecting line to generate multiple auxiliary points according to equation (3): （3）； in: Indicates the first The first column on the vertical connecting line Line auxiliary point coordinates, This indicates the first [unclear] generated by screen top border interpolation. Coordinates of auxiliary points Indicates the number of vertical grid lines. This indicates the first value generated by interpolation of the bottom border of the screen. Coordinates of auxiliary points.
5. The screen area adaptive geometric correction method according to claim 1, characterized in that: The method for mapping and aligning the two-dimensional source mesh with the transformed target image in step S4 to obtain the coordinate mapping relationship from the source space to the target space is as follows: For each horizontal line segment in the pixel grid of the target image, the corresponding coordinates of each pixel in the source space are calculated using the bilinear transformation algorithm according to equation (4), forming a pixel mapping relationship from the source space to the target space: （4）； in: This represents the horizontal line segment in each pixel grid of the target image being transformed. The corresponding coordinates of each pixel in the source space This represents the starting pixel of the horizontal line segment in each pixel grid of the target image being transformed. Indicates direction symbol, This represents the floor function. This represents the change in source space coordinates corresponding to one unit distance in the target space.
6. A screen region adaptive geometric correction system, used to perform a screen region adaptive geometric correction method as described in any one of claims 1 to 5, characterized in that: It includes an initial reference point acquisition module, an irregular outer border construction module, a two-dimensional source mesh construction module, a coordinate mapping relationship construction module, and a screen area adaptive geometric correction image generation module; The initial reference point acquisition module is used to select the four vertices of the effective display area of the source space screen image and the midpoint of the upper border as the initial reference points for describing the screen distortion contour. The irregular outer border construction module is used to generate multiple screen upper border auxiliary points using the initial reference point as the initial boundary control point and the boundary interpolation method to obtain an irregular screen upper border curve and a screen upper border point sequence. It also uses the linear interpolation method to generate multiple screen lower border auxiliary points to obtain an irregular screen lower border curve and a screen lower border point sequence. The two-dimensional source mesh construction module is used to connect the points at corresponding positions of the irregular screen top edge point sequence and the irregular screen bottom edge point sequence in pairs to form multiple vertical connection lines. Then, an interpolation algorithm is applied to each vertical connection line to generate multiple vertical connection line auxiliary points. The vertical connection line auxiliary points corresponding to different vertical connection lines are connected to form multiple horizontal connection lines. Finally, a two-dimensional source mesh is generated, which consists of an irregular screen top edge curve, an irregular screen bottom edge curve, multiple vertical connection lines, and multiple horizontal connection lines. The coordinate mapping relationship construction module is used to take a standard rectangular area corresponding to the screen area in the source space screen image as the transformation target image of the target space, and map and align the two-dimensional source grid with the transformation target image to obtain the coordinate mapping relationship from the source space to the target space. The screen region adaptive geometric correction image generation module is used to resample the source space screen image captured in real time by the camera. Based on the coordinate mapping relationship from the source space to the target space, it queries the coordinates of each pixel in the transformed target image in the source space screen image, and finally generates the screen region adaptive geometric correction image.