42 results about "Temporal filter" patented technology

In various examples, the present disclosure relates to using temporal filters for automated real-time classification. The technology described herein improves the performance of a multiclass classifier that may be used to classify a temporal sequence of input signals—such as input signals representative of video frames. A performance improvement may be achieved, at least in part, by applying a temporal filter to an output of the multiclass classifier. For example, the temporal filter may leverage classifications associated with preceding input signals to improve the final classification given to a subsequent signal. In some embodiments, the temporal filter may also use data from a confusion matrix to correct for the probable occurrence of certain types of classification errors. The temporal filter may be a linear filter, a nonlinear filter, an adaptive filter, and / or a statistical filter.

The invention provides an infrared image flicker processing method, system and device and a medium. The method comprises the steps of generating a blocking piece mean value based on blocking piece image data of continuous N frames obtained during single-point correction; processing the histograms of the two-time single-point correction gap images according to the separation blade mean value to generate a statistical histogram of the images; performing first filtering processing on the statistical histogram to generate a first filtering result of the histogram; after the first filtering result is calculated in an accumulated mode, a gray mapping table is calculated, second filtering processing is carried out, and a second filtering result of the gray mapping table is generated; and performing time domain filtering processing on the second filtering result to realize flicker processing of the infrared image. According to the infrared image flicker processing method, the flicker phenomenon of the infrared image can be effectively reduced, and the stability and the visual effect of the image are improved.

Systems and methods herein are for a video encoder to be associated with a temporal filter and a coding tree and that can perform a main pass for video encoding using individual video blocks towards prediction of at least one frame associated with the media stream, where the coding tree is associated with a lookahead pass, and where the temporal filter can enable denoising within the lookahead pass to reduce an effect of noise in one or more of motion estimation or mode selection of the video encoding.

In various embodiments, when reference picture resampling is applied to achieve better coding efficiency, methods and devices adapt temporal filtering of a noise-reduced input video prior to encoding according to a resolution ratio. According to one embodiment, temporal filtering strength values are fine-tuned at a pre-processing module according to the resolution ratio of the RPR mode. According to another embodiment, temporal filtering is deferred after rescaling when the RPR mode is enabled, and filtering strength values are fine-tuned according to the resolution ratio of the RPR mode. A third embodiment proposes to refine motion estimation in the encoder by setting the block size of motion estimation in temporal filtering according to the resolution ratio of the RPR mode.

Light time-of-flight camera for distance measurement with a time-of-flight sensor with at least one time-of-flight pixel, with a lighting system for emitting modulated light, the time-of-flight camera is designed for distance measurements according to a phase measurement principle and is designed in this way, that, a) for at least a part of the light-time-of-flight pixels, for determining a distance (d w ) of an object at least two measurements (A1, A2, ...) are carried out with different modulation frequencies (f1, f2, ...) and thus different uniqueness ranges (UR1, UR2, ...), whereby in each measurement (A i ) a residual distance (d i = mod(d, UR i )) is measured, b) starting from these residual distances (d i ) for each light travel time pixel, initially a sub-area index M of a total distance curve (UR) extd) is assigned, c) a field of the initially assigned sub-area indices M is processed using a "nearest neighbor filter" and / or a temporal filter, to generate a corrected, final sub-area index (M_final) for the respective light-time-of-flight pixel, and only then, d) the distance (d w ) by projecting the remaining distances (d i ) is determined on that sub-area of ​​the total distance curve which is defined by the corrected, final sub-area index (M_final).

A system and method for identifying container numbers on front-mounted cranes are disclosed, relating to the field of container number identification on front-mounted cranes. To address the shortcomings of existing methods, such as inability to adapt to rainy or snowy weather and low-light conditions, poor container number identification performance, and poor anti-interference capabilities, this invention acquires video or images of containers on front-mounted cranes; extracts continuous frames from the video at a fixed frequency and performs grayscale processing on each frame to reduce computational complexity; applies temporal filtering to the preprocessed video / image; applies spatial filtering to the filtered video / image to remove isolated noise points; reduces high-frequency noise and smooths the image through weighted averaging; enhances image contrast by adjusting the grayscale distribution of the image, making the container number area clearer; and performs container number area detection and character recognition on the processed image to achieve automatic container number identification. This invention is mainly used for identifying container numbers on front-mounted cranes.

The invention discloses a display device and a backlight control method thereof, the display device comprises a processor, a backlight module and a display module, the processor determines a first backlight value corresponding to a first backlight partition based on initial pixel values of a plurality of target pixel points; performing spatial filtering processing on the first backlight value to obtain a second backlight value; performing time filtering processing on the second backlight value based on the third backlight value and the filtering coefficient to obtain a target backlight value; the backlight module lights up a backlight source in the first backlight partition according to the target backlight value, and backlight is provided; the display module displays an image frame to be displayed. Through spatial filtering processing, the purpose of adjusting the luminance of the connecting position of every two adjacent backlight subareas is achieved, and the cross color phenomenon of a spatial domain is eliminated. The time filtering processing with different processing intensities is carried out, so that the lightening rates of the backlight sources in the backlight partitions are kept consistent, the cross color phenomenon of the time domain is eliminated, and the display effect of the display equipment is improved.

The present application discloses a video compression method, apparatus, device and storage medium, relating to the technical field of video processing, including: determining a target video sequence to be compressed, and sequentially obtaining a plurality of pixel blocks in each frame image of the target video sequence; respectively calculating the sum of squared errors, vertical sum of squared errors and horizontal sum of squared errors of the plurality of pixel blocks in each frame image; calculating the error distribution of the corresponding pixel blocks based on the size, sum of squared errors, vertical sum of squared errors and horizontal sum of squared errors of each pixel block to obtain an error distribution value, and determining whether the error distribution value is less than a preset threshold; if it is less than the preset threshold, performing a temporal filtering operation on the corresponding pixel block to obtain filtered data, and performing encoding and compression on the filtered data to obtain a compression result; if it is not less than the preset threshold, directly performing encoding and compression on the corresponding pixel block to obtain a compression result. The present application can improve the compression performance of the encoder and at the same time improve the efficiency of video compression.

The invention relates to adaptive quantization based on a coding tree. Systems and methods herein are used to associate a video encoder with a temporal filter and a coding tree, and may perform a main channel of video coding using respective video blocks to predict at least one frame associated with a media stream, where the coding tree is associated with a look-ahead channel, and the temporal filter may denoise within the look-ahead channel, and where the coding tree is associated with the look-ahead channel. To reduce noise effects in one or more of motion estimation or mode selection for video coding.

The invention discloses a bare soil intelligent identification method and system of a high-resolution remote sensing image, and relates to the technical field of ecological environment remote sensing monitoring, high-resolution remote sensing image data of a to-be-identified area is acquired and preprocessed, and preprocessed high-resolution remote sensing image data is obtained; performing filtering optimization on the preprocessed high-resolution remote sensing image data by adopting a dynamic adaptive multi-temporal filtering optimization method to obtain an optimized high-resolution remote sensing image data image; cutting the optimized high-resolution remote sensing image data, extracting contour position coordinates of a mask image, and constructing a training set; and inputting the training set into a bare soil recognition model for training to obtain a trained bare soil recognition model, and recognizing the bare soil plaques of the to-be-recognized region based on the trained bare soil recognition model. The multi-scale bare soil plaques under a complex background are accurately and automatically identified under a relatively low computing resource demand, and the distinction degree between the bare soil and other ground features is enhanced.

This paper presents a method for fusing and reconstructing trajectory data from multiple sources in hybrid traffic flow, relating to the fields of multi-source sensor fusion localization and target tracking. The method acquires the 3D acceleration sequence and RGB video stream of the target object, extracting the geometric center coordinate sequence. Based on the frequency domain energy distribution of acceleration, the target's fundamental frequency parameters are obtained. Temporal filtering coefficients are constructed using the dynamic sampling rate of the video, and the geometric center coordinates are filtered to obtain a corrected center coordinate sequence, which is then mapped to a measurement vector. Kinematic state prediction is performed using acceleration to obtain prior inferred state variables, and then posterior updates are performed using the measurement vectors to obtain the optimal spatial position. The reconstructed trajectory sequence is generated by stitching the data together, and the sliding time window is updated based on feedback-optimized parameters. This method effectively eliminates spurious displacements caused by gait, fundamentally eliminating jagged oscillations in the trajectory, resulting in a smooth and stable fused trajectory.

A neural network-based medical image processing apparatus according to the present invention, which identifies a small intestine region from a medical image acquired by a capsule endoscope, comprises: a memory equipped with an organ identification algorithm for performing organ identification from the medical image; and a processor for applying the medical image to the organ identification algorithm such that the small intestine region is identified. The organ identification algorithm includes: a convolutional neural network algorithm for identifying an organ included in the medical image as the stomach, the small intestine, and the large intestine to identify the small intestine region therefrom; and a temporal filtering algorithm linked to the convolutional neural network algorithm to reduce images which may be misidentified by the convolutional neural network algorithm.

Embodiments described relate to adjusting one or more pixel values in a region corresponding to an image artifact based on one or more spatial characteristics and one or more temporal characteristics to reduce or eliminate the image artifact. An electronic device may employ spatio-temporal filtering circuitry that includes spatial adjustment circuitry, temporal adjustment circuitry, and / or fuse circuitry. The spatial adjustment circuitry may perform a spatial adjustment of image data in the region of the image frame. Moreover, the temporal adjustment circuitry may perform a temporal adjustment of the image data in the region of the image frame. The fuse circuitry may then merge the spatially adjusted image data and the temporally adjusted image data in the region of the image frame.

The application discloses a training method and a decoding method of a single-time spatial-temporal filter based on double-frequency intermodulation, and the training method comprises the following steps: displaying a plurality of training targets in a row-column distribution on a visual stimulation interface, wherein each training target is jointly coded by a first stimulation signal and a second stimulation signal; collecting single-time training data of a subject; pre-processing the single-time training data to obtain ; performing time-domain displacement processing on the pre-processed mean template of the training target to obtain a time-shifted template; integrating the pre-processed mean template and the time-shifted template of the plurality of training targets to obtain a training target matrix; obtaining a standard reference signal of the training target, wherein the standard reference signal comprises a double-frequency intermodulation component; obtaining a training target reference signal matrix according to the standard reference signal; and obtaining a first spatial-temporal filter according to the training target reference signal matrix. The application achieves the technical effect that a spatial-temporal filter with good performance is obtained by using single-time training data, and provides a new method for efficient encoding and decoding of a large-instruction-set visual evoked brain-computer interface.

A decoder includes circuitry configured to receive a bitstream, decode a plurality of video frames from the bitstream, determine that a long-term reference block update mode is enabled for a current block of a current frame, determine that a long-term reference block update including pixel values and using the plurality of video frames, and determine that a long-term reference block update using the plurality of video frames is enabled for a current block of the current frame. And updating a portion of a long-term reference frame with the long-term reference block update. Related apparatus, systems, techniques, and articles are also described.

The present invention proposes a sparse image view synthesis method based on time-domain filtered neural radiation fields. This method innovatively introduces a time-domain filtered neural radiation field model to solve the problem that the neural radiation field represented by the multi-layer perceptron has high rendering overhead and cannot reduce the aliasing noise of the synthesized image by oversampling. By applying filtering technology in the time domain, the computational burden required for oversampling is effectively shared over consecutive frames, thereby significantly improving the quality of the synthesized image and effectively reducing noise and aliasing without increasing the number of additional network inferences. In addition, in order to improve the geometric extraction capability of the neural radiation field, the method introduces a deep prior network and designs an inter-frame geometric consistency loss to enhance the expression capability of the neural radiation field in multiple views and improve the generalization performance of the network.

The invention provides an image encoding method and an image encoding device. The image coding method comprises the steps that first quantization parameter information of a code rate control unit block in an image to be filtered is acquired, and the code rate control unit block represents a processing unit used for code rate distribution and image quality control; adjusting a filtering weight of a filtering block to be subjected to motion compensation time domain filtering in the bit rate control unit block based on the first quantization parameter information; performing motion compensation time domain filtering on the filtering block based on the adjusted filtering weight; and performing encoding based on the filtering result.

Provided is a method for encoding a video, an electronic device and a storage medium, relating to the field of video encoding. The method includes: obtaining a reference frame that is adjacent to a current frame and has established hierarchical division of an initial ROI; reusing a motion vector generated by the reference frame in a motion compensated temporal filter process to determine an ROI position prediction value of the current frame; extracting facial feature points from the current frame, and correcting the ROI position prediction value based on a spatial topological relationship of feature points in the initial ROI; configuring differentiated quantization parameters for a main face region, a secondary face region and a background region after correction respectively; dynamically allocating a three-region bitrate according to a real-time network bandwidth; and outputting an encoded frame of the current frame and associated ROI-level metadata.

There is provided a method, performed by a temporal filter. The method comprises obtaining a current picture and a set of one or more neighbor pictures comprising a first neighbor picture, wherein each of the one or more neighbor pictures either precedes or follows the current picture in a picture count order. The method comprises, based on a first block of pixel values of the current picture and a first block of pixel values in the first neighbor picture, determining a first filter weight, W1, for at least a first pixel value in the first block of pixel values of the current picture, wherein the first block of pixel values in the first neighbor picture correspond to the first block of pixel values in the current picture. The method comprises, calculating a first filtered pixel value, Pf1, at least using as inputs to the calculation: W1, P1, and Pr1, where PI is the first pixel value from the first block of pixel values of the current picture, and Pr1 is a pixel value from the first neighbor picture that corresponds to the first pixel value.

The present disclosure provides systems, methods, and devices for image signal processing that support improved temporal filtering of image frames to correct for motion distortion. In a first aspect, a method of image processing includes receiving image data including a first image frame. One or more of (i) a gain map of the first image frame and / or (ii) a motion map of the first image frame may be calculated and may be used to determine at least one parameter for the temporal filtering process. The corrected image frame may be determined by temporally filtering the first image frame based on the at least one parameter. Other aspects and features are also claimed and described.

Various approaches are disclosed to temporally and spatially filter noisy image data—generated using one or more ray-tracing effects—in a graphically rendered image. Rather than fully sampling data values using spatial filters, the data values may be sparsely sampled using filter taps within the spatial filters. To account for the sparse sampling, locations of filter taps may be jittered spatially and / or temporally. For filtering efficiency, a size of a spatial filter may be reduced when historical data values are used to temporally filter pixels. Further, data values filtered using a temporal filter may be clamped to avoid ghosting. For further filtering efficiency, a spatial filter may be applied as a separable filter in which the filtering for a filter direction may be performed over multiple iterations using reducing filter widths, decreasing the chance of visual artifacts when the spatial filter does not follow a true Gaussian distribution.

The present disclosure provides a video encoding method and apparatus, an electronic device, and a non-transitory computer-readable storage medium. The video encoding method includes performing motion estimation and motion compensation on a current block; calculating a block error between an extended block of the current block and an extended block of a motion compensated block; determining a filtering weight based on the block error; temporally filtering the current block by performing a weighted sum of the current block and the motion compensated block based on the determined filtering weight; and encoding the current block based on a result of the temporal filtering. The present disclosure can consider the relationship between a block-level filtering weight factor and surrounding pixels in a motion compensation-based temporal filtering process, and utilize the surrounding pixels to correct the result of the block-level error term in the temporal filtering weight, thereby improving the temporal filtering effect and improving the encoding efficiency.

Machine learning system and method and device for creating a machine learning system. The present invention relates to a machine learning system (10) comprising at least one temporal filter (21). Input variables comprising a temporal sequence of images are processed by means of the machine learning system (10) using the filter (21). The machine learning system (10) is configured to apply the filter (21) to a sequence of pixels that are all at the same coordinates of the image or at the same coordinates of an intermediate result. The filter coefficients of the filter (21) are quantized. The present invention also relates to a method, a computer program, and a device for creating a machine learning system (10).

Provided are a video coding method and apparatus, an electronic device, a program product, a storage medium, and a bit stream storage method, the video coding method comprising: performing motion compensation time domain filtering on a current frame based on the current frame and a reference frame of the current frame; the current frame is coded based on the motion compensation time domain filtering result of the current frame, and motion compensation time domain filtering is carried out on the current frame based on the current frame and the reference frame of the current frame, and comprises the following steps: in the process of carrying out motion compensation time domain filtering on the current frame, carrying out motion compensation time domain filtering on the reference frame of the current frame; spatial enhancement filtering is performed on the input of the motion compensation phase, and / or spatial enhancement filtering is performed on the input of the filtering phase.

This invention relates to the field of sound field reproduction technology, specifically, to a temporal broadband optimization algorithm for speaker placement in a personal audio zone system, comprising the following steps: assuming L0 speakers are located at L0 candidate positions, the l-th speaker is muted, and the temporal filter coefficients of the remaining L0-1 speakers are calculated; based on the obtained optimal speaker filter coefficients, the performance index of the remaining L0-1 speakers is calculated when the l-th speaker is muted; the performance index corresponding to l is calculated for l taking values ​​of 1, 2, ..., L0, and the speaker position corresponding to the minimum performance index is found and denoted as l0; the l0-th speaker is removed from the candidate positions, and the remaining L0-1 speaker positions are taken as new candidate positions; for the new L0-1 speaker candidate positions, steps one to three are repeated for iterative optimization until the pre-set L speaker positions are selected.

A system for regulating temporal filtering strength is configurable to: (i) obtain a light level indicator indicating light level associated with a real-world environment; (ii) determine a motion compensation confidence indicator using a current image and a previous image; (iii) determine a filter weight by processing the light level indicator and the motion compensation confidence indicator using a filter strength regulation module; and (iv) generate an output image by using at least the filter weight to filter a current frame with a previous frame.

The invention belongs to the technical field of computer vision and video analysis, and particularly relates to a time sequence adaptive filtering weak supervision video anomaly detection method based on CLIP assistance. According to the method, a time sequence modeling and filtering framework fusing vision-language cross-modal information is designed, and the method comprises the steps that firstly, video local and global time sequence dependence is decomposed and modeled through a local-global time sequence adapter, and the capturing capacity of a model for the dynamic evolution process of abnormal events is enhanced; secondly, a double-branch architecture is adopted, pure visual features are used for completing coarse-grained anomaly detection, fine-grained anomaly recognition is achieved in combination with visual-language alignment features, and detection requirements of different complexity scenes are met; and finally, introducing a Fourier transform time sequence filter, adaptively suppressing noise in a frequency domain, strengthening abnormal feature representation, and improving the anti-interference capability of the model. The method is remarkably superior to an existing weak supervision method, and the precision and robustness of anomaly detection can be effectively improved through the synergistic effect of the modules.

A method for a multi-antenna receiver arrangement, the multi-antenna receiver arrangement being comprisable in a wireless device, WD, the method comprising: receiving a first plurality of analog radio signals; converting the first plurality of analog radio signals into a first plurality of digital signals; extracting reference signals from each of the first plurality of digital signals; determining characteristics for each of the first plurality of digital signals based on the extracted reference signals; processing the first plurality of digital signals with a second plurality of spatial-temporal filters to obtain a second plurality of combined signals, wherein the first plurality is larger than the second plurality. Corresponding computer program product, multi-antenna receiver arrangement, and wireless device are also disclosed.

Embodiments described relate to adjusting one or more pixel values in a region (90) corresponding to an image artifact (80) based on one or more spatial characteristics and one or more temporal characteristics to reduce or eliminate the image artifact (80). An electronic device (10) may employ spatio-temporal filtering circuitry (110) that includes spatial adjustment circuitry (130), temporal adjustment circuitry (132), and / or fuse circuitry (134). The spatial adjustment circuitry (130) may perform a spatial adjustment of image data in the region (90) of the image frame. Moreover, the temporal adjustment circuitry (132) may perform a temporal adjustment of the image data in the region (90) of the image frame. The fuse circuitry (134) may then merge the spatially adjusted image data and the temporally adjusted image data in the region (90) of the image frame.

In video compression, the quantization parameter (QP) determines the compression level and bitrate of videos. Block-level QP adaptation can improve video encode efficiency, but determining suitable QPs at the block-level without incurring significant overhead is not trivial. To address this concern, QP adaptation at a QP unit having one or more blocks can be performed using one or more filtering distortion values which are already calculated for the one or more blocks by a motion compensated temporal filter (MCTF). One or more minimal distortion values are used in calculating a distortion metric, which is then used for determining the quantization parameter to be applied to a QP unit. The one or more minimal distortion values are also used for determining a size of the QP unit. Different thresholds may be applied to the distortion metric, depending on whether MCTF is on or off.