MMVD MERGE MODE WITH ASYMMETRIC MOTION VECTOR DIFFERENCE

    公开(公告)号:US20230370605A1

    公开(公告)日:2023-11-16

    申请号:US18305807

    申请日:2023-04-24

    CPC classification number: H04N19/137 H04N19/105 H04N19/176 H04N19/70

    Abstract: An example device for decoding video data includes one or more processors configured to determine merge mode information for a current block, the merge mode information indicating that motion information for a current block is to be predicted using a first predictor motion vector and a second predictor motion vector; determine a first motion vector difference (MVD) for the first predictor motion vector and a second MVD for the second predictor motion vector, the second MVD being different than the first MVD; form a first motion vector equaling a combination of the first motion vector predictor and the first MVD; form a second motion vector equaling a combination of the second motion vector predictor and the second MVD; generate a prediction block using the first motion vector and the second motion vector; and decode the current block using the prediction block.

    Adaptive loop filter with fixed filters

    公开(公告)号:US11778177B2

    公开(公告)日:2023-10-03

    申请号:US17557706

    申请日:2021-12-21

    Abstract: A video decoder can be configured to apply a first stage adaptive loop filter (ALF) to a reconstructed sample by determining a first class index for the reconstructed sample, selecting a filter from a first set of filters based on the first class index, and applying the filter from the first set of filters to the reconstructed sample to determine a first intermediate sample value; apply a second stage ALF to the reconstructed sample by determining a second class index for the reconstructed sample; select a second filter from a second set of filters based on the second class index, applying the second filter to the reconstructed sample to determine a first sample modification value, and determining a second sample modification value based on the first intermediate sample value; and determine a filtered reconstructed sample based on the reconstructed sample and the first and second sample modification values.

    TEMPORAL INITIALIZATION POINTS FOR CONTEXT-BASED ARITHMETIC CODING

    公开(公告)号:US20230283782A1

    公开(公告)日:2023-09-07

    申请号:US18176863

    申请日:2023-03-01

    CPC classification number: H04N19/152 H04N19/423 H04N19/174 H04N19/172

    Abstract: A method includes determining one or more context values for at least one context used for encoding or decoding a current slice or picture, determining that a buffer for storing sets of temporal initialization points from two or more slices or pictures for context-based arithmetic coding is full, determining a first set of temporal initialization points associated with a slice or picture, from among the two or more slices or pictures, based on at least one of a slice type, a temporal identification value, or a quantization parameter (QP) value of the slice or picture, removing the first set of temporal initialization points that is associated with the slice or picture, and storing a second set of temporal initialization points associated with the current slice or picture, wherein the second set of temporal initialization points are based on the determined one or more context values.

Patent Agency Ranking