Image prediction method and apparatus performing intra prediction
An image decoding method according to the present document comprises the steps of: generating a prediction sample for the current block on the basis of an intra prediction mode for the current block and neighboring reference samples; and generating a reconstructed picture for the current block on the basis of the prediction sample, wherein, when the reference sample position indicated by a prediction direction of a directional prediction mode of the current block from the position of the prediction sample is the fractional sample position, the prediction sample is derived from among the neighboring reference samples by interpolating with respect to specific neighboring reference samples located around the fractional sample position, when the specific neighboring reference samples include outer neighboring reference samples located at (2*nW+1, −1) and (2*nW+2, −1) positions, the values of the outer neighboring reference samples are set to be equal to the value of a sample located at (2*nW, −1), and nW may indicate the width of the current block.
1. An image decoding method performed by a decoding apparatus, the method comprising:
deriving an intra prediction mode for a current block;
deriving neighboring reference samples of the current block;
generating a prediction sample for the current block based on the intra prediction mode and the neighboring reference samples; and
generating a reconstructed picture for the current block based on the prediction sample,
wherein the intra prediction mode for the current block is a directional prediction mode;
wherein based on a position indicated by a prediction direction of the directional prediction mode with regard to a position of the prediction sample being a position of a fractional sample, the prediction sample is derived based on interpolation of specific neighboring reference samples located neighboring the position of the fractional sample,
wherein the specific neighboring reference samples are included in the neighboring reference samples,
wherein the interpolation of the specific neighboring reference samples is performed based on an interpolation filter,
wherein based on the specific neighboring reference samples comprising an outer neighboring reference sample located to a right of a position of (2W−1, −1) or located below a position of (−1, 2H−1), a reconstructed sample value of the outer neighboring reference sample is used for the interpolation,
wherein W is a width of the current block and H is a height of the current block,
wherein the neighboring reference samples include top neighboring reference samples, a position of a right-most reference sample among the top neighboring reference samples is determined based on a filter tap size of the interpolation filter, and
wherein the position of the right-most reference sample is (2W+N−1, −1), wherein N is determined based on a following equation:
N =(filterSize>>1)−1
where filterSize represents the filter tap size and >> represents arithmetic right shift.
2. The image decoding method of claim 1 , wherein the filter tap size is 4, and
wherein the interpolation filter includes a Gaussian filter or a cubic filter.
3. The image decoding method of claim 1 , wherein based on the neighboring reference samples forming a plurality of reference sample lines, the reconstructed sample value of the outer neighboring reference sample is derived based on reference sample line index information indicating one of the plurality of the reference sample lines.
4. The image decoding method of claim 1 , wherein based on a position of a top-left sample position in the current block being (0, 0), a position of the outer neighboring reference sample is (2W, −1) or (−1, 2H).
5. An image encoding method performed by an encoding apparatus, the method comprising:
deriving an intra prediction mode for a current block;
deriving neighboring reference samples of the current block;
generating a prediction sample for the current block based on the intra prediction mode and the neighboring reference samples;
generating information on the intra prediction mode; and
encoding the information on the intra prediction mode,
wherein the intra prediction mode for the current block is a directional prediction mode;
wherein based on a position indicated by a prediction direction of the directional prediction mode with regard to a position of the prediction sample being a position of a fractional sample, the prediction sample is derived based on interpolation of specific neighboring reference samples located neighboring the position of the fractional sample,
wherein the specific neighboring reference samples are included in the neighboring reference samples,
wherein the interpolation of the specific neighboring reference samples is performed based on an interpolation filter,
wherein based on the specific neighboring reference samples comprising an outer neighboring reference sample located to a right of a position of (2W−1, −1) or located below a position of (−1, 2H−1), a reconstructed sample value of the outer neighboring reference sample is used for the interpolation,
wherein W is a width of the current block and H is a height of the current block,
wherein the neighboring reference samples include top neighboring reference samples, a position of a right-most reference sample among the top neighboring reference samples is determined based on a filter tap size of the interpolation filter, and
wherein the position of the right-most reference sample is (2W+N−1, −1), wherein N is determined based on a following equation:
N =(filterSize>>1)−1
where filterSize represents the filter tap size and >> represents arithmetic right shift.
6. The image encoding method of claim 5 , wherein the filter tap size is 4, and
wherein the interpolation filter includes a Gaussian filter or a cubic filter.
7. The image encoding method of claim 5 , wherein based on the neighboring reference samples forming a plurality of reference sample lines, the reconstructed sample value of the outer neighboring reference sample is derived based on reference sample line index information indicating one of the plurality of the reference sample lines.
8. The image encoding method of claim 5 , wherein based on a position of a top-left sample position in the current block being (0, 0), a position of the outer neighboring reference sample is (2W, −1) or (−1, 2H).
9. A non-transitory computer-readable storage medium storing a bitstream generated by a method, the method comprising:
deriving an intra prediction mode for a current block;
deriving neighboring reference samples of the current block;
generating a prediction sample for the current block based on the intra prediction mode and the neighboring reference samples; and
generating the bitstream by encoding information on the intra prediction mode,
wherein the intra prediction mode for the current block is a directional prediction mode;
wherein based on a position indicated by a prediction direction of the directional prediction mode with regard to a position of the prediction sample being a position of a fractional sample, the prediction sample is derived based on interpolation of specific neighboring reference samples located neighboring the position of the fractional sample,
wherein the specific neighboring reference samples are included in the neighboring reference samples,
wherein the interpolation of the specific neighboring reference samples is performed based on an interpolation filter,
wherein based on the specific neighboring reference samples comprising an outer neighboring reference sample located to a right of a position of (2W−1, −1) or located below a position of (−1, 2H−1), a reconstructed sample value of the outer neighboring reference sample is used for the interpolation,
wherein W is a width of the current block and H is a height of the current block,
wherein the neighboring reference samples include top neighboring reference samples, a position of a right-most reference sample among the top neighboring reference samples is determined based on a filter tap size of the interpolation filter, and
wherein the position of the right-most reference sample is (2W+N−1, −1), wherein N is determined based on a following equation:
N =(filterSize>>1)−1
where filterSize represents the filter tap size and >> represents arithmetic right shift.
10. The non-transitory computer-readable storage medium of claim 9 , wherein based on a position of a top-left sample position in the current block being (0, 0), a position of the outer neighboring reference sample is (2W, −1) or (−1, 2H).