Inter prediction mode-based image processing method and apparatus therefor
Disclosed are an inter prediction mode-based image processing method and an apparatus therefor. Particularly, a method for processing an image on the basis of inter prediction may comprise the steps of: determining whether a motion vector scale adaptation is applied to a block; up-scaling a down-scaled MVD (Motion Vector Difference) when the motion vector scale adaptation is applied to the block; deriving a MV (Motion Vector) for the block, using the up-scaled MVD and a MVP (Motion Vector Predictor); and generating a predictive block of the block, using the derived MV.
1. A method of decoding a video signal, comprising:
obtaining residual samples of a current block;
deriving a motion vector predictor (MVP) based on motion information of a neighbor block of the current block;
obtaining a motion vector difference (MVD) of the current block;
obtaining a flag related to a resolution of the MVD, a value 0 of the flag specifying that the resolution of the MVD is ¼ pel-scale;
rounding the MVP based on the flag;
scaling the MVD based on the flag;
obtaining a motion vector (MV) of the current block based on the scaled MVD and the rounded MVP;
generating prediction samples of the current block based on the MV; and
generating reconstructed samples of the current block based on the prediction samples and the residual samples.
2. The method of claim 1 , wherein the MVD is obtained based on a plurality of syntax elements, comprising:
a first flag including information about whether an absolute value of the MVD is greater than 0;
a second flag including information about an absolute value of the MVD is greater than 1, the second flag being parsed when the MVD is greater than 0;
a value obtained by subtracting 2 from an absolute value of the MVD; and
a third flag including information about a sign of the MVD.
3. The method of claim 2 , wherein the second flag and the third flag are parsed when the MVD is greater than 0, and the value is parsed when the second flag is not equal to 0.
4. The method of claim 1 , wherein the scaling of the MVD up-scales the MVD.
5. The method of claim 1 , wherein the flag is parsed when the obtained MVD is not 0.
6. The method of claim 1 , wherein a round function is applied to the MVP and the MVD is multiplied by 4 based on that the flag is equal to a first value.
7. The method of claim 1 , wherein the MVD is grouped in units of intervals and binary-encoded.
8. The method of claim 7 , wherein the MVD is encoded as a start value of an interval to which the MVD belongs and indication information for indicating the MVD within the interval.
9. The method of claim 8 , wherein the start value is down-scaled at a ratio of a size of the interval and transmitted.
10. The method of claim 9 , wherein the size of the interval is predetermined or transmitted from an encoder.
11. The method of claim 7 , wherein the MVD is grouped in units of intervals and encoded only when the MVD is 2 or more.
12. The method of claim 7 wherein a size of the interval is set to a square number of 2.
13. An apparatus for decoding a video signal, comprising:
a memory configured to store the video signal; and
a processor coupled to the memory and configured to:
obtain residual samples of a current block;
derive a motion vector predictor (MVP) based on motion information of a neighbor block of the current block;
obtain a motion vector difference (MVD) of the current block;
obtain a flag related to a resolution of the MVD, a value 0 of the flag specifying that the resolution of the MVD is ¼ pa-scale;
round the MVP based on the flag;
scaling the MVD based on the flag;
obtain a motion vector (MV) of the current block based on the scaled MVD and the rounded MVP;
generate prediction samples of the current block based on the MV; and
generate reconstructed samples of the current block based on the prediction samples and the residual samples.