Encoder-side search ranges having horizontal bias or vertical bias
Innovations in encoder-side search ranges having horizontal bias or vertical bias are described herein. For example, a video encoder determines a block vector (“BV”) for a current block of a picture, performs intra prediction for the current block using the BV, and encodes the BV. The BV indicates a displacement to a region within the picture. When determining the BV, the encoder checks a constraint that the region is within a BV search range having a horizontal bias or vertical bias. The encoder can select the BV search range from among multiple available BV search ranges, e.g., depending at least in part on BV values of one or more previous blocks, which can be tracked in a histogram data structure.
1. In a computer system that implements a video encoder, a method comprising:
determining a block vector (“BV”) value for a current block of a given coding tree block of a current picture, the BV value indicating a displacement to a region within the current picture, wherein the determining the BV value for the current block includes identifying the BV value for the current block subject to a constraint that the region is within a BV search range having a horizontal bias or a vertical bias, wherein, for the horizontal bias, the BV search range includes candidate BV values having a wider range of horizontal BV component values than vertical BV component values, and wherein, for the vertical bias, the BV search range includes candidate BV values having a wider range of vertical BV component values than horizontal BV component values;
performing intra block copy prediction for the current block using the BV value for the current block; and
encoding the BV value for the current block.
2. The method of claim 1 , wherein the BV search range has a vertical bias.
3. The method of claim 2 , wherein the given coding tree block has a dimension S, and wherein the BV search range has a height between S and 2S, inclusive.
4. The method of claim 3 , wherein the BV search range has a width between ¼S and ¾S, inclusive.
5. The method of claim 1 , wherein the BV search range has a horizontal bias.
6. The method of claim 5 , wherein the given coding tree block has a dimension S, and wherein the BV search range has a width between S and 2S, inclusive.
7. The method of claim 6 , wherein the BV search range has a height between ¼S and ¾S, inclusive.
8. The method of claim 1 , further comprising selecting the BV search range from among multiple available BV search ranges.
9. The method of claim 8 , wherein the selecting:
depends at least in part on BV values of one or more previous blocks of the current picture and/or one or more previous pictures; or
depends at least in part on BV values of one or more previous blocks, wherein the BV values of the one or more previous blocks are tracked in a histogram data structure.
10. The method of claim 8 , wherein the selecting depends at least in part on a user setting.
11. The method of claim 1 , wherein a 2N×2N block includes the current block, the method further comprising identifying how to partition the 2N×2N block using a bottom-up approach that includes:
checking a subset of modes per N×N block of the 2N×2N block;
caching vector values for the respective N×N blocks of the 2N×2N block;
checking a subset of modes with a 2N-dimension for the 2N×2N block, including using the cached vector values;
selecting a best mode with a 2N-dimension for the 2N×2N block; and
selecting between the best mode with a 2N-dimension for the 2N×2N block and best modes for the respective N×N blocks of the 2N×2N block.
12. The method of claim 1 , wherein the determining the BV value for the current block includes:
identifying a current best location for prediction through iterative evaluation in a small neighborhood around the current best location, wherein the small neighborhood includes locations that are immediately adjacent horizontally or vertically to the current best location; and
confirming the current best location through iterative evaluation in successively larger neighborhoods around the current best location, wherein each of the larger neighborhoods includes locations in a ring outside the small neighborhood.
13. The method of claim 1 , wherein the current block is part of a current coding tree block having dimensions N×N, the method further comprising identifying how to partition the current coding tree block into two partitions that have different dimensions.
14. The method of claim 1 , wherein the current block is part of a current slice of the current picture, and wherein the identifying the BV value for the current block is also subject to a constraint that the region is within the current slice.
15. A computer system comprising one or more processing units and memory, wherein the computer system implements a video decoder configured to perform operations comprising:
receiving encoded data, as part of a bitstream, for a current picture of a video sequence; and
decoding the current picture using the encoded data, wherein the decoding the current picture includes:
decoding a block vector (“BV”) value for a current block of a given coding tree block of the current picture, the BV value indicating a displacement to a region within the current picture, wherein the BV value for the current block is subject to a constraint that the region is within a BV search range having a horizontal bias or a vertical bias, wherein, for the horizontal bias, the BV search range includes candidate BV values having a wider range of horizontal BV component values than vertical BV component values, and wherein, for the vertical bias, the BV search range includes candidate BV values having a wider range of vertical BV component values than horizontal BV component values; and
performing intra block copy prediction for the current block using the BV value for the current block.
16. The computer system of claim 15 , wherein the BV search range has a horizontal bias, wherein the given coding tree block has a dimension S, and wherein the BV search range has a width between S and 2S, inclusive.
17. The computer system of claim 15 , wherein the current block is part of a current slice of the current picture, the BV value for the current block is also subject to a constraint that the region is within the current slice.
18. One or more non-transitory computer-readable media having stored thereon encoded data, in a bitstream, for a current picture of a video sequence, the encoded data including data representing a block vector (“BV”) value for a current block of a given coding tree block of the current picture, the BV value indicating a displacement to a region within the current picture, wherein the BV value for the current block is subject to a constraint that the region is within a BV search range having a horizontal bias or a vertical bias, wherein, for the horizontal bias, the BV search range includes candidate BV values having a wider range of horizontal BV component values than vertical BV component values, and wherein, for the vertical bias, the BV search range includes candidate BV values having a wider range of vertical BV component values than horizontal BV component values, the encoded data resulting from encoding of the current picture by operations that include:
determining the BV value for the current block, including identifying the BV value for the current block subject to the constraint that the region is within the BV search range having the horizontal bias or the vertical bias;
performing intra block copy prediction for the current block using the BV value for the current block; and
encoding the BV value for the current block.
19. The one or more computer-readable media of claim 18 , wherein the BV search range has a horizontal bias, wherein the given coding tree block has a dimension S, and wherein the BV search range has a width between S and 2S, inclusive.
20. The one or more computer-readable media of claim 18 , wherein the current block is part of a current slice of the current picture, and wherein the BV value for the current block is also subject to a constraint that the region is within the current slice.