Video decoding apparatus
An image decoding apparatus is implemented that can suppress a decrease in coding efficiency in a case that a high compression rate is achieved. The image decoding apparatus includes a parameter decoder, and the parameter decoder decodes a skip flag indicating whether a skip mode is applied, and in a case that the skip flag does not indicate the skip mode, decodes a merge flag indicating whether a merge mode is applied, and in a case that the merge flag does not indicate the merge mode, decodes an MMVD flag indicating whether an MMVD mode is applied.
1. A video decoding apparatus comprising,
a parameter decoding circuit configured to decode parameters for generating a prediction image in an MMVD mode; and
an MMVD prediction circuit configured to obtain motion vectors by adding an L0 vector difference or an L1 vector difference to motion vectors for merging,
wherein the parameter decoding circuit decodes a first index specifying an index in a merge candidate list, a second index specifying a length of the L0 vector difference or the L1 vector difference, and a third index specifying a direction of the L0 vector difference or the L1 the vector difference,
the MMVD prediction circuit
derives the motion vectors for merging from the first index,
derives a first vector from the second index and the third index,
in a case that a distance between a target picture and a L0 picture is greater than a distance between the target picture and a L1 picture, sets the L0 vector difference equal to the first vector, and derives the L1 vector difference by scaling the first vector,
in a case that a distance between a target picture and a L0 picture is less than a distance between a target picture and an L1 picture, sets the L1 vector difference equal to the first vector, and derives the L0 vector difference by scaling the first vector.
2. The video decoding apparatus according to the claim 1 ,
wherein, in a case that a distance between a target picture and the L0 picture is equal to a distance between the target picture and the L1 picture, and both the L0 picture and the L1 picture are located in same direction against the target picture in time order, the MMVD prediction circuit sets the L0 vector difference and the L1 vector difference equal to the first vector.
3. A video decoding method comprising,
decoding parameters for generating a prediction image in an MMVD mode;
decoding a first index specifying an index in a merge candidate list, a second index specifying a length of an L0 vector difference or an L1 vector difference, and a third index specifying a direction of the L0 vector difference of the L1 vector difference;
deriving motion vectors for merging from the first index;
deriving a first vector from the second index and the third index;
in a case that a distance between a target picture and a L0 picture is greater than a distance between the target picture and a L1 picture, setting the L0 vector difference equal to the first vector, and deriving the L1 vector difference by scaling the first vector;
in a case that the distance between the target picture and the L0 picture is less than the distance between the target picture and the L1 picture, setting the L1 vector difference equal to the first vector, and deriving the L0 vector difference by scaling the first vector; and
obtaining motion vectors motion vectors by adding the L0 vector difference or the L1 vector difference to the motion vectors for merging.
4. The video decoding apparatus according to the claim 1 ,
wherein, in a case that the distance between the target picture and the L0 picture is equal to the distance between the target picture and the L1 picture, and the L0 picture, the target picture, and the L1 picture are arranged in time order, the MMVD prediction circuit sets the L0 vector difference equal to the first vector, and the L1 vector difference is set equal to a result of multiplying the first vector by −1.
5. A video coding apparatus comprising:
a parameter coding circuit configured to code parameters for generating a prediction image in an MMVD mode; and
an MMVD prediction circuit configured to obtain motion vectors by adding an L0 vector difference or an L1 vector difference to motion vectors for merging; wherein
the parameter coding circuit codes a first index specifying an index in a merge candidate list, a second index specifying a length of the L0 vector difference or the L1 vector difference, and a third index specifying a direction of the L0 vector difference or the L1 vector difference; and
the MMVD prediction circuit:
derives the motion vectors for merging from the first index;
derives a first vector from the second index and the third index;
in a case that a distance between a target picture and a L0 picture is greater than a distance between the target picture and an L1 picture, sets the L0 vector difference equal to the first vector, and derives the L1 vector difference by scaling the first vector; and
in a case that the distance between the target picture and the L0 picture is less than the distance between the target picture and the L1 picture, sets the L1 vector difference equal to the first vector, and derives the L0 vector difference by scaling the first vector.