Method and apparatus for video coding using implicit arbitrary block partition and prediction according thereto

1 . A method performed by a video decoding apparatus for intra-predicting a current block, the method comprising:

selecting a reference region of the current block, the reference region being a previously reconstructed region that is present in at least one of a top or a left of the current block;

deriving arbitrary partitioning information of the current block implicitly by using directionality information derived from the reference region;

partitioning the current block into sub-regions according to the arbitrary partitioning information;

determining a prediction mode for each sub-region of the sub-regions;

generating predicted signals of each sub-region by using the prediction mode; and

generating and using weights to perform weighted summing on the predicted signals of the sub-regions to generate final predicted signals of the current block.

2 . The method of claim 1 , wherein deriving the arbitrary partitioning information comprises:

dividing the reference region into reference subblocks that are non-overlapping and have a predetermined size.

3 . The method of claim 2 , wherein deriving the arbitrary partitioning information further comprises:

calculating a pixel value mean of pixel values included in each reference subblock for each of the reference subblocks;

at the top of the current block, selecting at least one pair of adjacent reference subblocks with a greater disparity between pixel value means than a preset mean disparity threshold;

at the left of the current block, selecting at least one pair of adjacent reference subblocks with a greater disparity between pixel value means than the preset mean disparity threshold; and

deriving the arbitrary partitioning information based on a straight line that connects a boundary between the pair of the adjacent reference subblocks selected at the top of the current block with a boundary between the pair of the adjacent reference subblocks selected at the left of the current block.

4 . The method of claim 2 , wherein deriving the arbitrary partitioning information further comprises:

calculating a variance of pixel values included in each reference subblock for each of the reference subblocks to select at least one reference subblock with a greater variance than a preset variance threshold; and

deriving the arbitrary partitioning information by extending, over the current block, a straight line which represents a directionality within the selected reference subblocks.

5 . The method of claim 2 , wherein deriving the arbitrary partitioning information further comprises:

calculating a variance of pixel values included in each reference subblock for each of the reference subblocks to select a reference subblock that has a largest variance; and

deriving the arbitrary partitioning information by extending, over the current block, a straight line which represents a directionality within the selected reference subblock.

6 . The method of claim 5 , wherein deriving the arbitrary partitioning information further comprises:

when the reference subblock with the largest variance exists at the top of the current block and has the directionality of a horizontal direction, or when the reference subblock with the largest variance exists on the left of the current block and has the directionality of a vertical direction, deriving the arbitrary partitioning information by extending, over the current block, a straight line which represents a directionality in a reference subblock that has a second largest variance.

7 . The method of claim 2 , wherein deriving the arbitrary partitioning information further comprises:

detecting a pair of adjacent reference subblocks where strong filtering is applied on a boundary between the adjacent reference subblocks during an in-loop filtering process; and

then deriving the arbitrary partitioning information by extending, over the current block, a straight line which represents a directionality within a block of a predetermined size that partially overlaps the adjacent reference subblocks and includes the boundary between the adjacent reference subblocks.

8 . The method of claim 2 , wherein determining the prediction mode comprises:

utilizing a most probable mode (MPM) list for determining the prediction mode for each sub-region of the sub-regions, the MPM list being constructed by using prediction modes of previously reconstructed reference subblocks neighboring the sub-regions.

9 . The method of claim 4 , wherein determining the prediction mode comprises:

decoding, from a bitstream, a flag that indicates whether to use a prediction mode having a directionality that coincides with a directionality used to partition the sub-regions.

10 . The method of claim 9 , wherein determining the prediction mode further comprises:

when the flag is true, determining a prediction mode that has a directionality used for partitioning each sub-region of some sub-regions, as a prediction mode for each sub-region.

11 . The method of claim 1 , wherein weighted summing the predicted signals of the sub-regions comprises:

deriving the weights implicitly by using an amount of change in luminance values of predicted signals of each sub-region relative to a partitioning boundary which divides the current block into the sub-regions.

12 . The method of claim 1 , further comprising:

decoding, from a bitstream, an arbitrary split flag that indicates whether the current block is to be partitioned based on an arbitrary partitioning; and

checking for the arbitrary split flag,

wherein when the arbitrary split flag is true, selecting the reference region of the current block.

13 . A method performed by a video encoding apparatus for intra-predicting a current block, the method comprising:

selecting a reference region of the current block, the reference region being a previously reconstructed region that is present in at least one of a top or a left of the current block;

deriving arbitrary partitioning information of the current block implicitly by using directionality information derived from the reference region;

partitioning the current block into sub-regions according to the arbitrary partitioning information;

determining a prediction mode for each sub-region of the sub-regions;

generating predicted signals of each sub-region by using the prediction mode; and

generating and using weights to perform weighted summing on the predicted signals of the sub-regions to generate final predicted signals of the current block.

14 . The method of claim 13 , further comprising:

determining an arbitrary split flag that indicates whether the current block is to be partitioned based on an arbitrary partitioning; and

checking for the arbitrary split flag,

wherein when the arbitrary split flag is true, selecting the reference region of the current block.

15 . The method of claim 14 , further comprising:

encoding the arbitrary split flag.

16 . A method for providing a video decoding apparatus with video data, the method comprising:

encoding the video data into a bitstream; and

transmitting the bitstream to the video decoding device,

wherein encoding the video data comprises:

selecting a reference region of a current block, the reference region being a previously reconstructed region that is present in at least one of a top or a left of the current block;

deriving arbitrary partitioning information of the current block implicitly by using directionality information derived from the reference region;

partitioning the current block into sub-regions according to the arbitrary partitioning information;

determining a prediction mode for each sub-region of the sub-regions;

generating predicted signals of each sub-region by using the prediction mode; and

generating and using weights to perform weighted summing on the predicted signals of the sub-regions to generate final predicted signals of the current block.