Interframe prediction processor with address management mechanism for motion vector storage
A device for processing variable-size macroblocks to achieve high-speed interframe prediction. A predictor calculator determines a motion vector predictor (MVP) for a current macroblock from motion vectors of other macroblocks adjacent to the current macroblock. To achieve this, the predictor calculator specifies adjacent macroblocks to an address calculator. The address calculator calculates a vector memory address for each specified adjacent macroblock, based on the block size of that specified macroblock. The predictor calculator reads out motion vectors from the received addresses and determines MVP for the current macroblock.
1. An interframe prediction processor that performs interframe prediction of video frames on a block basis, comprising:
a reference picture memory providing vector storage locations corresponding to minimum size blocks to store motion vectors calculated for blocks of variable sizes;
an address calculator that calculates an address of a vector storage location in the reference picture memory, based at least on the size of a specified block;
a predictor calculator that specifies blocks adjacent to a current block to the address calculator so as to obtain therefrom addresses of motion vectors calculated for the adjacent blocks specified, reads out the motion vectors from the obtained addresses of the reference picture memory, and determines a motion vector predictor of the current block based on the motion vectors read out of the reference picture memory; and
a reference picture storage manager that writes the calculated motion vectors in the vector storage locations reserved in the reference picture memory;
wherein the address calculator determines the address of the vector storage location, based on the block size of the specified adjacent block, when an address request specifying an adjacent block is received from the predictor calculator;
and wherein the address calculator comprises a translation table that associates each combination of block size and block location with a specific address of the reference picture memory; and the address calculator uses the translation table to determine the address of a motion vector in the reference picture memory, based on the block size and block location of the specified block.
2. The interframe prediction processor according to claim 1 , wherein:
the vector storage locations in the reference picture memory include current-block vector storage locations for storing motion vectors of blocks within the current block and adjacent-block vector storage locations for storing motion vectors of blocks adjacent to the current block; and
the reference picture storage manager copies the calculated motion vector from the current-block vector storage locations to the adjacent-block vector storage locations before a block located next to the current block is selected as a new current block.
3. The interframe prediction processor according to claim 2 , wherein the address calculator comprises:
a first translation table associating each combination of the block size and block location with an address of one of the current-block vector storage locations, and
a second translation table associating each combination of the block size and block location with an address of one of the adjacent-block vector storage locations,
and wherein the address calculator uses the first and second translation tables to determine the address of a vector storage location corresponding to the specified block.
4. A method of performing interframe prediction of video frames on a block basis, comprising the steps of:
providing vector storage locations corresponding to minimum size blocks to store motion vectors calculated for blocks of variable sizes;
specifying blocks adjacent to a current block;
calculating addresses of vector storage locations corresponding to the specified adjacent blocks, based at least on the size of each of the specified adjacent blocks;
reading out motion vectors from the calculated addresses;
determining a motion vector predictor of the current block based on the motion vectors that are read out;
writing a calculated motion vector in the vector storage locations;
determining the addresses of the vector storage locations, based on the block size of the specified adjacent block, when an address request specifying an adjacent block is received;
associating, in a translation table, each combination of block size and block location with a specific address of the reference picture memory; and
determining the address of a motion vector in the reference picture memory, based on the block size and block location of the specified block in the translation table.
5. A video coding device that encodes video frames into compressed video data by performing motion-compensated prediction between a source picture and a reference picture on a block basis, comprising:
a reference picture memory providing vector storage locations corresponding to minimum size blocks to store motion vectors calculated for blocks of variable sizes;
an address calculator that calculates an address of a vector storage location in the reference picture memory, based at least on the size of a specified block;
a predictor calculator that specifies blocks adjacent to a current block to the address calculator so as to obtain therefrom addresses of motion vectors calculated for the adjacent blocks specified, reads out the motion vectors from the obtained addresses of the reference picture memory, and determines a motion vector predictor of the current block based on the motion vectors read out of the reference picture memory; and
a reference picture storage manager that writes a calculated motion vector in the vector storage locations reserved in the reference picture memory;
wherein the address calculator determines the address of the vector storage location, based on the block size of the specified adjacent block, when an address request specifying an adjacent block is received from the predictor calculator;
and wherein the address calculator comprises a translation table that associates each combination of block size and block location with a specific address of the reference picture memory; and the address calculator uses the translation table to determine the address of a motion vector in the reference picture memory, based on the block size and block location of the specified block.
6. A video decoding device that reconstructs original video frames from a compressed video signal that has been produced by performing motion-compensated interframe prediction on a block basis, comprising:
a reference picture memory providing vector storage locations corresponding to minimum size blocks to store motion vectors calculated for blocks of variable sizes;
an address calculator that calculates an address of a vector storage location in the reference picture memory, based at least on the size of a specified block;
a predictor calculator that specifies blocks adjacent to a current block to the address calculator so as to obtain therefrom addresses of motion vectors calculated for the adjacent blocks specified, reads out the motion vectors from the obtained addresses of the reference picture memory, determines a motion vector predictor of the current block based on the motion vectors read out of the reference picture memory, and reconstructs a motion vector of the current block from the determined motion vector predictor and a decoded motion vector difference; and
a reference picture storage manager that writes the reconstructed motion vector in the vector storage locations reserved in the reference picture memory;
wherein the address calculator determines the address of the vector storage location, based on the block size of the specified adjacent block, when an address request specifying an adjacent block is received from the predictor calculator;
and wherein the address calculator comprises a translation table that associates each combination of block size and block location with a specific address of the reference picture memory; and the address calculator uses the translation table to determine the address of a motion vector in the reference picture memory, based on the block size and block location of the specified block.