MOTION VECTOR CODING USING DYNAMIC REFERENCE MOTION VECTORS

    公开(公告)号:US20170324978A1

    公开(公告)日:2017-11-09

    申请号:US15147053

    申请日:2016-05-05

    Applicant: GOOGLE INC.

    Abstract: Techniques are described to code motion vectors using reference motion vectors to reduce the amount of bits needed. One method includes determining, for a current block of the video bitstream, a reference motion vector from a varying number of candidate reference motion vectors, wherein the reference motion vector is associated with a reference block and includes a predicted portion and a residual portion; selecting a probability context model for the current block by evaluating the residual portion of the reference motion vector with one or more thresholds; and entropy decoding, for the current block using a processor, a motion vector residual associated with the current block using the probability context model.

    ENTROPY CODING TRANSFORM PARTITIONING INFORMATION

    公开(公告)号:US20170237980A1

    公开(公告)日:2017-08-17

    申请号:US15044233

    申请日:2016-02-16

    Applicant: Google Inc.

    Abstract: Entropy coding of transform partitioning information may include decoding a current block by determining a probability for entropy decoding a transform partitioning flag based on a transform block size of an adjacent block that is available for decoding, entropy decoding the transform partitioning flag for the current block using the probability, generating a reconstructed block based on the current block by, on a condition that the transform partitioning flag indicates that transform partitioning is omitted for the current block, inverse transforming the current block using a current block size inverse transform, and, on a condition that the transform partitioning flag indicates transform partitioning for the current block, identifying sub-blocks from the current block, wherein each sub-block from the sub-blocks has a sub-block size smaller than the current block size, inverse transforming the sub-blocks using sub-block size inverse transforms, and generating the reconstructed block based on the residual block.

    Adaptive motion search control for variable block size partitions in video coding

    公开(公告)号:US09967558B1

    公开(公告)日:2018-05-08

    申请号:US14108773

    申请日:2013-12-17

    Applicant: Google Inc.

    CPC classification number: H04N19/00733 H04N19/119 H04N19/147 H04N19/176

    Abstract: A block of a video data stream may be subject to a motion search for comparing inter prediction of the blocks to intra prediction of the block while being partitioned. A motion search using an initial search range finds a first motion vector for the block. The blocks is partitioned into at least two sub-blocks, each having a same size, and one or more motion searches are performed for each sub-block using a second search range and the first motion vector where the second search range is a fraction of the initial search range. Respective error values associated with each of the motion searches are compared to select a partition for coding the block, and the block may be encoded using the selected partition. Performing inter prediction in this fashion can improve the performance of an encoder in terms of both execution time and compression ratio.

    Entropy coding transform partitioning information

    公开(公告)号:US09942548B2

    公开(公告)日:2018-04-10

    申请号:US15044233

    申请日:2016-02-16

    Applicant: Google Inc.

    Abstract: Entropy coding of transform partitioning information may include decoding a current block by determining a probability for entropy decoding a transform partitioning flag based on a transform block size of an adjacent block that is available for decoding, entropy decoding the transform partitioning flag for the current block using the probability, generating a reconstructed block based on the current block by, on a condition that the transform partitioning flag indicates that transform partitioning is omitted for the current block, inverse transforming the current block using a current block size inverse transform, and, on a condition that the transform partitioning flag indicates transform partitioning for the current block, identifying sub-blocks from the current block, wherein each sub-block from the sub-blocks has a sub-block size smaller than the current block size, inverse transforming the sub-blocks using sub-block size inverse transforms, and generating the reconstructed block based on the residual block.

    CONTEXT ADAPTIVE SCAN ORDER FOR ENTROPY CODING

    公开(公告)号:US20180070106A1

    公开(公告)日:2018-03-08

    申请号:US15259189

    申请日:2016-09-08

    Applicant: GOOGLE INC.

    Abstract: Video coding using a context adaptive scan order for entropy coding may include decoding a current block by identifying a context adaptive scan order for entropy decoding a transform block, which may include identifying non-zero-coefficient probabilities for the transform block such that each location in the transform block corresponds to a respective non-zero-coefficient probability from the non-zero-coefficient probabilities, assigning a respective context adaptive scan order position to each location in the transform block in descending magnitude order of the respective corresponding non-zero-coefficient probabilities such that the context adaptive scan order position for each location exceeds the context adaptive scan order position assigned to entropy coding context locations for the respective location, entropy decoding transform coefficients from the encoded video stream based on the context adaptive scan order, and reconstructing the decoded block based on the transform block.

Patent Agency Ranking