A video predictive decoding method executed by a video predictive decoding device, comprising:
- decoding side information and a compressed data of a residual signal of a target region, which is a target to be decoded, out of plural sets of compressed data obtained by encoding a plurality of divided regions;
- restoring a motion vector used to generate a predicted signal of the target region from the side information;
- storing the motion vector in a motion information storing unit;
- generating the predicted signal of the target region, based on the motion vector;
- restoring a reproduced residual signal of the target region from the compressed data comprising the residual signal;
- and
- adding the predicted signal to the reproduced residual signal to reconstruct a pixel signal of the target region, and storing the reconstructed pixel signal as a previously-reproduced picture,
- wherein decoding side information comprises decoding zero-th side information and first side information,
- wherein the zero-th side information contains a zero-th motion vector difference, and a zero-th motion vector predictor index used to identify as a zero-th motion vector predictor one motion vector selected from a plurality of motion vectors stored in the motion information storing unit,
- wherein the first side information contains a first motion vector predictor index used to identify as a first motion vector predictor one motion vector selected from a plurality of motion vectors stored in the motion information storing unit,
- wherein restoring a motion vector comprises:
- deriving the zero-th motion vector predictor based on the zero-th motion vector predictor index by using the one motion vector selected from the motion vectors stored in the motion information storing unit, and adding the zero-th motion vector predictor to the zero-th motion vector difference to reconstruct a zero-th motion vector;
- and
- setting 0 for a vector value of a first motion vector difference, deriving the first motion vector predictor based on the first motion vector predictor index by using the one motion vector selected from the motion vectors stored in the motion information storing unit, and adding the first motion vector predictor to the first motion vector difference to reconstruct a first motion vector, and
- wherein the zero-th side information further contains a zero-th reference frame index to identify a zero-th reference frame for generating a zero-th predicted signal of the target block, and when the zero-th reference frame identified by the zero-th reference frame index is different from a reference frame identified by the zero-th motion vector predictor index, a scaling process of the zero-th motion vector predictor is performed based on frame numbers of the decoding target frame, the zero-th reference frame of the target block, and the reference frame identified by the zero-th motion vector predictor index before adding the zero-th motion vector predictor to the zero-th motion vector difference to reconstruct the zero-th motion vector, and
- wherein the first side information further contains a first reference frame index to identify a first reference frame for generating a first predicted signal of the target block, and when the first reference frame identified by the first reference frame index is different from a reference frame identified by the first motion vector predictor index, a scaling process of the first motion vector predictor is performed based on frame numbers of the decoding target frame, the first reference frame of the target block, and the reference frame identified by the first motion vector predictor index before adding the first motion vector predictor to the first motion vector difference to reconstruct the first motion vector,
- wherein generating the predicted signal of the target region comprises combining two signals acquired from the previously-reproduced picture, based on the zero-th motion vector and the first motion vector respectively, to generate the predicted signal of the target region.