19 results about "Sum of absolute differences" patented technology

The invention relates to an image matching method, an image detection device and a storage medium, and the method comprises the steps: obtaining a template image and a to-be-detected image, constructing a first image of an image pyramid of the template image, and constructing a second image of the image pyramid of the to-be-detected image, matching the first image and the second image of the top layer based on a preset square difference and an algorithm to obtain a first matching position in the second image of the top layer, and mapping each first matching position as a matching initial position of a secondary top layer according to a pyramid mapping relation, and matching the first image and the second image of the second top layer based on a preset absolute value error and an algorithm until the matching of the bottom layer is completed, and obtaining an accurate matching area corresponding to each to-be-searched target in the to-be-detected image. According to the method, different algorithms are used for performing matching processing on different layers of images of the pyramid, and the calculation complexity is reduced on the premise of ensuring the matching alignment accuracy, so that the template matching efficiency is effectively improved.

An image decoding method comprises the steps of: determining whether or not an application condition of decoder-side motion vector refinement (DMVR) for applying motion vector refinement for a current block is satisfied; deriving a minimum sum of absolute differences (SAD) on the basis of L0 and L1 motion vectors of the current block if the application condition of the DMVR is satisfied; deriving refined L0 and L1 motion vectors of the current block based on the minimum SAD; deriving prediction samples of the current block based on the refined L0 and L1 motion vectors; and generating reconstructed samples of the current block based on the prediction samples. With respect to whether or not the application condition of the DMVR is satisfied, the application condition of the DMVR is determined to be satisfied if a prediction mode, in which inter-prediction and intra-prediction are combined, is not applied to the current block.

The invention provides a semi-global stereo matching method based on dynamic cross aggregation and multidirectional scanning line optimization, and relates to the technical field of computer vision and three-dimensional reconstruction. According to the method, the initial matching cost is constructed by fusing the pixel absolute difference sum matching cost and the sorting transformation matching cost, the dynamic cross cost aggregation region is constructed adaptively based on the color difference and the spatial distance, and on this basis, semi-global cost optimization is executed by adopting four-way scanning line iteration to generate the disparity map. And further combining left and right consistency check, region voting update, shielding filling, edge constrained neural network residual correction and sub-pixel fitting to output an optimized disparity map, thereby improving the disparity matching stability and overall consistency of a weak texture region and a depth discontinuous region.

The invention provides a motion texture joint threshold (WTJT) method, which comprises the following steps of: firstly, comparing the motion speed and the motion vector of a current frame with the motion speed and the motion vector of a previous frame and a next frame through the sum of absolute differences (SAD) in a frame interpolation information acquisition mode, and screening candidate jump blocks; secondly, extracting two textural features of direction and frequency details from adjacent frames based on a local binary pattern (LBP), and screening candidate hop blocks again in combination with Gabor filtering; and finally, performing spatial-temporal feature decomposition on an input video sequence, establishing a proportional relation between motion information and texture feature information, and calculating a joint weighting coefficient (JWC) so as to dynamically adjust a block jumping threshold value of each frame and realize adaptive allocation of a block jumping proportion. Through united weighting after motion and texture features are normalized, the method provided by the invention improves the accuracy of block jumping judgment, optimizes the distribution of block jumping rate, and achieves balance between algorithm complexity and reconstruction quality.

A method of processing an indexed group of images, the method comprises identifying a first group of images, such as a first texture atlas, and a second group of images; identifying a first image, e.g. a first texture patch, in the first group of images 61; determining a second image, in the second group of images 62, the second image being similar to the first image; and updating an index of the second image based on an index of the first image 63. The second group of images is associated with a three-dimensional representation of a scene, wherein texture points in the three-dimensional represent groups of reference images or points within the second group. The index represents relative position and attributes of points in the image. Assessing similarities comprises determining a distance between the first image and the second image; a Euclidean distance therebetween or a sum of absolute differences (SAD). The process may be iterative and before identifying the first image, reordering of the first and second groups of images based on luminance is carried out. Indices may be arranged in a z-pattern. (Fig 14a)

An image decoding method performed by a decoding apparatus comprises obtaining information on an inter prediction mode for a first target block; deriving motion information of the first target block based on the information on the inter prediction mode; deriving modified motion information of the first target block by searching for position with a minimum sum of absolute differences (SAD) value within a predetermined area, based on the motion information of the first target block and bi-prediction; configuring a merge candidate list for a second target block to which a merge mode is applied, based on a spatial neighboring block and a temporal neighboring block of the second target block; obtaining a merge index and residual information for the second target block; selecting one of merge candidates constituting the merge candidate list based on the merge index for the second target block; deriving motion information of the second target block based on the selected merge candidate; generating prediction samples for the second target block by performing inter prediction based on the motion information of the second target block; generating residual samples for the second target block based on the residual information; and generating reconstructed samples for the second target block based on the prediction samples and the residual samples. The merge candidates include a spatial motion information candidate and a temporal motion information candidate. The spatial motion information candidate is derived based on the spatial neighboring block and the temporal motion information candidate is derived based on the temporal neighboring block. In a case where the temporal neighboring block is the first target block, the temporal motion information candidate is derived based on the modified motion information of the first target block rather than the motion information of the first target block. In a case where the spatial neighboring block is the first target block, the spatial motion information candidate is derived based on the motion information of the first target block rather than the modified motion information of the first target block.

The present disclosure relates to an adaptive semi-global stereo matching parameter setting method, device, electronic equipment and storage medium. The method comprises: based on a left camera image of a binocular camera shooting a target, obtaining the coordinates of a preset left target area and calculating the image proportion; calculating the global variance and the local variance; calculating the local inverse luminosity according to the local variance, generating the minimum disparity number based on a preset correspondence relationship; calculating the joint inverse luminosity according to the image proportion, the global variance and the local variance, generating the sum of absolute differences cost calculation window based on a preset correspondence relationship; configuring the minimum disparity number and the sum of absolute differences cost calculation window to complete the adaptive setting of the semi-global stereo matching parameters. By calculating the minimum disparity number and the sum of absolute differences cost calculation window size of the image, the present disclosure can adapt to various lighting conditions and perform autonomous parameter setting under non-human control.

Methods and apparatuses of determining an alignment level between motion compensated reference patches for reducing motion vector refinement steps are provided. One method implemented by a decoder includes, obtaining motion compensated interpolated samples based on sub-pixel accurate merge motion vectors from a bilinear motion compensated interpolation, computing a sum of absolute differences (SAD) between two motion compensated reference patches using a subset of the motion compensated interpolated samples, and determining whether the SAD is less than a coding unit (CU) size-dependent threshold value. The method further includes when the SAD is less than the CU size-dependent threshold value, skipping remaining decoder-side motion vector refinement (DMVR) process steps and performing final motion compensation. When the SAD is not less than the CU size-dependent threshold value, performing the remaining DMVR process steps and performing the final motion compensation.

A method of block matching between first data and second data comprises populating a buffer in internal memory of a processor with initial summed area table (SAT) data based on a difference between at least an initial portion of the first data and second data. An iterative process is performed comprising selecting a first portion of the first data and corresponding second portion of the second data using a sliding window which moves with a predefined window offset for each iteration. Subsets of the first and second portion are determined and absolute difference (AD) data, e.g. sum of absolute differences (SAD) data, between the subsets is calculated. SAT data is computed using the AD data and at least a portion of the SAT data stored in the buffer. The SAT data in the buffer is then updated by replacing a subset of it with the newly calculated SAT data. A plurality of comparisons between the first and second data are performed and difference data is calculated for each comparison using the SAT data in the buffer. The difference data is used to determine matched blocks. Buffering the SAT data reduces redundant calculations.

According to aspects of the invention there is provided a method of modifying sets of residuals data where residual data can be used to correct or enhance data of a base stream, for example a frame of a video encoded using a legacy video coding technology. According to a first aspect there is provided a method of encoding an input video, the method comprising: receiving an input video comprising a plurality of frames; generating one or more sets of residuals on a frame-by-frame basis based on a difference between a given frame of the input signal and one or more reconstructed versions of the given frame at one or more respective spatial resolutions; selectively modifying the one or more sets of residuals; and encoding the one or more sets of modified residuals to generate one or more respective encoded streams for the input video, wherein the method further comprises: determining a sum of absolute differences pixel metric based on one or more of the plurality of frames, the pixel metric being determined for one or more coding units of the one or more of the plurality of frames; and wherein the one or more sets of residuals are selectively modified based on the pixel metric.

An image decoding method comprises the steps of: determining whether or not an application condition of decoder-side motion vector refinement (DMVR) for applying motion vector refinement for a current block is satisfied; deriving a minimum sum of absolute differences (SAD) on the basis of L0 and L1 motion vectors of the current block if the application condition of the DMVR is satisfied; deriving refined L0 and L1 motion vectors of the current block based on the minimum SAD; deriving prediction samples of the current block based on the refined L0 and L1 motion vectors; and generating reconstructed samples of the current block based on the prediction samples. With respect to whether or not the application condition of the DMVR is satisfied, the application condition of the DMVR is determined to be satisfied if a prediction mode, in which inter-prediction and intra-prediction are combined, is not applied to the current block.

The application discloses a kind of multi-view point cloud matching feature histogram construction method based on structured light phase information, by introducing phase information as auxiliary feature, to enhance the feature distinguishing ability of fast point feature histogram (FPFH), first by structured light technology obtains phase information, combined with traditional fast point feature histogram algorithm, the extended fast point feature histogram feature descriptor including normal vector angle, Euclidean distance and phase value is constructed.Normal vector angle and distance are quantified to N interval respectively, and phase information is introduced as the second 2N+1 feature dimension, the distinguishing degree and robustness of feature description are significantly improved.In the feature matching process, pixel-level matching is carried out using polar correction and sum of absolute differences (SAD) algorithm, and sub-pixel optimization is carried out combined with the linear characteristics of phase information, and the matching accuracy is improved.The application is excellent in the point cloud registration of weak texture area and flat surface, effectively improves the quality and accuracy of three-dimensional reconstruction under complex scene.

Disclosed are a method and an apparatus for encoding / decoding, in which the encoding method comprises performing motion compensation and local illumination compensation on a reference block by using a linear model, and in which a parameter of the linear model is determined using a sum of absolute differences of adjacent reconstructed samples and corresponding reference samples.

The invention provides a method for identifying similarity between two audio files or tracks. The method comprises receiving a processed audio file and an original audio file, uncompressing the processed audio file, applying global loudness normalization and short-term loudness normalization on the processed audio file and the original audio file, converting the processed audio file and the original audio file into processed spectral image by time-frequency mapping, scaling, using linear interpolation, the processed spectral image, dividing the scaled-up processed spectral image into slices, searching for minimum Sum of Absolute Difference (SAD), using original spectral image as reference, for each slice.

Encoding and decoding a current block of a current frame are disclosed. A first prediction block for the current block is obtained based on motion information associated with the first prediction block. A second prediction block for at least a portion of the current block is obtained based on motion information associated with the neighboring block. A prediction difference metric between the first prediction block and the second prediction block is obtained. The prediction difference metric is used to determine whether to combine the first prediction block and the second prediction block of the portion of the current block. The prediction difference metric may be a sum of absolute differences (SAD) between the first prediction block and the second prediction block. The prediction difference metric may be an absolute maximum of a pairwise difference between the first prediction block and the second prediction block.

Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. The method includes: determining, for a conversion between a video unit of a video and a bitstream of the video, a prediction of the video unit, a transformed sum of absolute difference (SATD) cost for a one-dimensional array or block with one-dimension array associated with the video unit; determining coding information of the video unit based on the SATD cost; and performing the conversion based on the coding information.

An image decoding method according to the present document comprises the steps of: determining whether or not an application condition of decoder-side motion vector refinement (DMVR) for applying motion vector refinement for a current block is satisfied; deriving a minimum sum of absolute differences (SAD) on the basis of L0 and L1 motion vectors of the current block if the application condition of the DMVR is satisfied; deriving refined L0 and L1 motion vectors of the current block on the basis of the minimum SAD; deriving prediction samples of the current block on the basis of the refined L0 and L1 motion vectors; and generating reconstructed samples of the current block on the basis of the prediction samples. With respect to whether or not the application condition of the DMVR is satisfied, the application condition of the DMVR is determined to be satisfied if a prediction mode, in which inter-prediction and intra-prediction are combined, is not applied to the current block.

Methods and apparatuses of determining an alignment level between motion compensated reference patches for reducing motion vector refinement steps are provided. According to one method, obtaining, by a decoder, motion compensated interpolated samples based on sub-pixel accurate merge motion vectors from a bilinear motion compensated interpolation; computing, by the decoder, a sum of absolute differences (SAD) between two motion compensated reference patches using a subset of the motion compensated interpolated samples; determining, by the decoder, whether the SAD is less than a coding unit (CU) size-dependent threshold value; when the SAD is less than the CU size-dependent threshold value: skipping remaining decoder-side motion vector refinement (DMVR) process steps; and performing final motion compensation; and when the SAD is not less than the CU size-dependent threshold value: performing the remaining DMVR process steps; and performing the final motion compensation.

Overlapped block motion compensation (OBMC) may be performed for a current video block based on motion information associated with the current video block and motion information associated with one or more neighboring blocks of the current video block. Under certain conditions, some or all of these neighboring blocks may be omitted from the OBMC operation of the current block. For instance, a neighboring block may be skipped during the OBMC operation if the current video block and the neighboring block are both uni-directionally or bi-directionally predicted, if the motion vectors associated with the current block and the neighboring block refer to a same reference picture, and if a sum of absolute differences between those motion vectors is smaller than a threshold value. Further, OBMC may be conducted in conjunction with regular motion compensation and may use simplified filters than traditionally allowed.