Abstract: A multi-stage shot encoding pipeline can be made more efficient by having the first-stage or first-pass encoder encode only the relevant frames of a shot, rather than all of the video frames of the shot. Reducing the video frames of the shot that needs to be processed by the first-stage or first-pass encoder can significantly reduce the processing time and storage requirements. Relevant frames can be identified using spatial and temporal characteristics of the video frames in each shot and select only the video frames that have the most relevant information. The spatial and temporal characteristics are already readily available in shot detection, making the solution almost cost-free to implement.

Abstract: An encoder may be tuned to handle video where most of the motion is horizontal. Re-tuning an encoder to content with non-horizontal motion is a difficult process. Instead of re-tuning the process, a solution for coding video content may include rotating a video before encoding it and rotating it back after decoding its compressed bitstream. This solution can allow encoders to consider the suitable orientation of each shot and apply different encoding parameters based on whether the shot needs to be rotated. Similarly, the solution may allow decoders to correctly display the video by rotating the decoded video data to the original orientation. Rotation information may be written into the encoded video bitstream, such as in a sequence header Open Bitstream Unit.

Abstract: In the process of block-prediction in block-based video compression, a vector for translating a reference block to produce a predicted block is not encoded directly in an encoded bitstream. Rather, a residual vector, which is the difference between the vector and a selected vector predictor candidate, is encoded to achieve higher compression efficiency. A selected vector predictor candidate can have the smallest Euclidean distance to the vector to ensure the residual vector is small. For compressing video that has complex spatial and/or temporal characteristics, the selected vector predictor candidate may not result in the smallest residual vector. To address this concern, vector predictor candidates are selected separately for the horizontal component and the vertical component of the vector to obtain smaller residual vectors. An effective and efficient signaling scheme can be implemented to indicate whether the predictor is based on components from two different predictor candidates.

Abstract: In block-based video compression, interpolation filters may be used in the process of motion compensation or in block-prediction to blend the pixels in the predicted block spatially and generate the reconstructed block. However, the content within a block can vary significantly across the block and using a single interpolation filter type for the entire block may not be sufficient to provide effective motion compensation. To address the concern more effectively, a subblock adaptive interpolation filtering approach can be implemented in a video codec to improve the quality of the reconstructed block while being able to keep file sizes small. Subblock adaptive interpolation filtering can be implemented by using different interpolation filter types for each subblock of the block. Subblock adaptive interpolation filtering can result in improved motion compensation and higher video quality.

Abstract: Techniques related to transform coefficient shaping for video encoding are discussed. Such techniques include applying weighting parameters from one or more perceptually-designed matrices of weighting parameters to blocks of transform coefficients to generate weighted transform coefficients and encoding the weighted transform coefficients into a bitstream. The process may be based on sets of perceptually designed matrices of weighting parameters. Classifier outputs may be used to select from the set of perceptually designed matrices a subset of matrices to work with. The latter may be used in a synthesis procedure to develop the final weighting matrix to be used is shaping the transform coefficients.

Abstract: Techniques related to video encoding are discussed that, for each block of input video, select an individual partitioning and coding mode selection technique from multiple such selection techniques. For a picture, the selection algorithm takes as input scores for individual blocks, costs of the various partitioning and coding mode selection techniques, and various detector outputs. The selection algorithm provides as output a partitioning and coding mode selection technique for each block in picture. The algorithms selection is such that the overall cost of the selected algorithms in the picture is as close as possible to a given picture budget. Furthermore, a partitioning and coding mode selection algorithms, binary depth partitioning (BDP), is discussed. For a block, BDP provides fast convergence to a partitioning and associated coding modes first evaluating intermediate partitioning options and converging on the final partitioning by evaluating either larger of smaller partitions.

Abstract: Techniques related to video encoding that provide for a decoupled prediction and coding structure for improved performance are discussed. Such techniques include determining final partitioning decisions for blocks of a picture by evaluating intra modes for candidate partitions by comparing the candidate partitions to intra predicted partitions generated using only original pixel samples and evaluating inter modes for the candidate partitions by comparing the candidate partitions to search partitions including original pixel samples and encoding using the final partitioning decision.

Abstract: Techniques related to detection of features and modification of encoding based on such detected features for improved data utilization efficiency are discussed. Such techniques include generating a partitioning decision for a block and coding mode decisions for partitions of the individual block using the detected features or indicators thereof based on one or more of generating a luma and chroma or luma only evaluation decision for a partition, generating a merge or skip mode decision for a partition having an initial merge mode decision, generating only a portion of a transform coefficient block for a partition, and evaluating 4×4 partitions only for any partition of the partitions that are 8×8 initial coding partitions.

Abstract: Techniques related to video encoding that provide for a decoupled prediction and coding structure for improved performance are discussed. Such techniques include determining final partitioning decisions for blocks of a picture by evaluating intra modes for candidate partitions by comparing the candidate partitions to intra predicted partitions generated using only original pixel samples and evaluating inter modes for the candidate partitions by comparing the candidate partitions to search partitions including original pixel samples and encoding using the final partitioning decision.

Abstract: Techniques related to transform coefficient shaping for video encoding are discussed. Such techniques include applying weighting parameters from one or more perceptually-designed matrices of weighting parameters to blocks of transform coefficients to generate weighted transform coefficients and encoding the weighted transform coefficients into a bitstream. The process may be based on sets of perceptually designed matrices of weighting parameters. Classifier outputs may be used to select from the set of perceptually designed matrices a subset of matrices to work with. The latter may be used in a synthesis procedure to develop the final weighting matrix to be used is shaping the transform coefficients.

Abstract: Techniques related to video encoding are discussed that, for each block of input video, select an individual partitioning and coding mode selection technique from multiple such selection techniques. For a picture, the selection algorithm takes as input scores for individual blocks, costs of the various partitioning and coding mode selection techniques, and various detector outputs. The selection algorithm provides as output a partitioning and coding mode selection technique for each block in picture. The algorithms selection is such that the overall cost of the selected algorithms in the picture is as close as possible to a given picture budget. Furthermore, a partitioning and coding mode selection algorithms, binary depth partitioning (BDP), is discussed. For a block, BDP provides fast convergence to a partitioning and associated coding modes first evaluating intermediate partitioning options and converging on the final partitioning by evaluating either larger of smaller partitions.

Abstract: Disclosed are adaptive loop filtering techniques for video encoding and/or decoding. For a video unit, the encoder selects a set of predefined filters or generates a set of new filters, and places into the bitstream information identifying the set of predefined filters, or information defining the set of new filters. The set of filters may be used for loop filtering of at least one of the reconstructed samples of the video unit. At the decoder, a set of filters may be obtained by, decoding an index that identifies a set of predefined filters, or by decoding information related to a set of new filters. The obtained set of filters may be used for loop filtering of at least one decoded and reconstructed sample of the video unit.

Abstract: Disclosed are adaptive loop filtering techniques in the context of video encoding and/or decoding. For each video unit, the encoder can select a filter shape, and can place into the bitstream information that identifies the filter shape. At least one filter whose shape is the selected filter shape is used to loop filter at least one sample. At the decoder, a filter shape is obtained by decoding information that identifies the filter shape. At least one filter whose shape is the obtained filter shape is used to loop filter at least one reconstructed sample. Different filter shapes are also disclosed.

Abstract: Disclosed are techniques for adaptive interpolation filtering of luminance and chrominance samples in the context of motion compensation in video encoding or decoding. A two-dimensional interpolation filter of n×m coefficients may be separable, i.e., it may be separated into two one-dimensional filters with m and n coefficients, respectively. The bitstream may include, per video unit and sub-sample position, information indicating whether to use a newly-generated, a cached, or a default filter that may be a separable two-dimensional filter. The information may be structured in a way that takes advantage of the two-dimensional filter being separable. When a newly-generated filter is signalled, the bitstream may contain information pertaining to the characteristics of the newly-generated filter, such as its coefficients. A decoder may fetch this information from the bitstream to create the filters which are applied to samples of the video unit. An encoder may create a bitstream as described.