Method and device for determination of motion vectors that are coordinated with regions of an image
View Patent ↗In a method and device for determining motion vectors that are each coordinated with individual regions of an image, the image is divided into a number of first main blocks, making use of a first block raster, and a first estimation process is performed to provide each first block with a motion vector. The image is also divided into a number of second main blocks, making use of a second block raster, whose position is displaced relative to or offset from the first block raster, for example in the horizontal and vertical directions, and a second estimation process is performed to provide each second main block with a motion vector. A number of subblocks are formed, each subblock being formed by an overlap region between one of the main blocks of the first raster and certain ones of the second main blocks of the second raster. A motion vector for each subblock is generated, this motion vector depending on the motion vectors of the main blocks and being coordinated with the particular subblock.
1. A method for determining motion vectors that are coordinated with individual image regions of an image, comprising:
dividing the image into a number of first main blocks using a first block raster and performing a first estimation process to provide a motion vector for each of the first main blocks;
dividing the image into a number of second main blocks using at least one second block raster, a position of the at least one second block raster being shifted relative to a position of the first block raster, and performing a second estimation process to provide a motion vector for each of the second main blocks;
forming a number of subblocks, each of the subblocks being formed by an intersection, region between one of the first main blocks and at least one of the second main blocks; and
generating a motion vector for each subblock as a function of the estimated motion vectors of the first and second main blocks whose intersection forms that subblock, each of the generated motion vectors for the subblocks being coordinated with the particular subblock;
where the dividing the image into the number of first main blocks, the dividing the image into the number of second main blocks, the forming a number of subblocks, and the generating the motion vector are performed by a video image processing device.
2. The method of claim 1 , where the dimensions of the first and second main blocks are equal.
3. The method of claim 2 , where the position of the second raster relative to the position of the first raster in the horizontal direction of the image constitutes half of the dimension of the main blocks in the horizontal direction and where the position of the second raster relative to the position of the first raster in the vertical direction of the image constitutes half of the dimension of the main blocks in the vertical direction.
4. The method of claim 1 , where the step of generating a motion vector for each subblock selects the motion vector of one of the main blocks that is part of the overlap intersection, region used to form a subblock as the motion vector of the subblock.
5. The method of claim 4 , further comprising the step of determining a measure of quality for each one of the motion vectors of the main blocks, where the step of generating the motion vector for each subblock selects the motion vector of one of the main blocks as the motion vector for the subblock depending on the determined main block motion vector quality measure.
6. The method claim 5 , further comprising the step of determining a measure of quality for the motion vector of each one of the subblocks, where the step of generating a motion vector for each subblock selects the motion vector of one of the main blocks as the motion vector of the subblock depending on the determined subblock motion vector quality measure.
7. The method of claim 4 , further comprising the step of determining at least one convergence criterion for each one of the motion vectors of the main blocks, where the step of generating a motion vector for each subblock selects the motion vector of one of the main blocks as the motion vector for the subblock depending on the at least one determined convergence criterion.
8. The method of claim 7 , where the step of determining at least one convergence criterion compares the motion vector of a first one of the main blocks to at least one motion vector of a second one of the main blocks that is adjoining the first main block and is in the same raster as the raster of the first one of the main blocks.
9. The method of claim 7 , where the step of determining at least one convergence criterion determines at least one convergence criterion for one of the main blocks of the first raster by comparing the motion vector of the one main block of the first raster to the motion vector of a second main block adjoining the left side of the one main block and to the motion vector of a third main block adjoining the top of the one main block, and where the step of determining at least one convergence criterion determines at least one convergence criterion for one of the main blocks of the second raster by comparing the motion vector of the one main block of the second raster to the motion vector of a second main block adjoining the right side of the one main block and to the motion vector of a third main block adjoining the bottom of the one main block.
10. The method of claim 1 , further comprising the step of filtering to select vector components of the motion vector of the subblock from corresponding vector components of the motion vectors of the main blocks that are coordinated with the subblock.
11. The method of claim 10 , where the step of filtering comprises the step of filtering by an ordered Statistical filter.
12. The method of claim 4 , further comprising the steps of:
forming the vector absolute values of the motion vectors of the main blocks coordinated with the subblock;
filtering the vector absolute values to provide one of the vector absolute values as a filter;
selecting as the motion vector of the subblock the motion vector of the main block whose vector absolute value was provided as the filter output.
13. The method of claim 12 , where the step of filtering comprises the step of filtering by an ordered Statistical filter.
14. A device for determining motion vectors that are coordinated with individual regions of an image, comprising:
means for dividing the image into a number of first main blocks using a first block raster and means for performing a first estimation process to provide a motion vector for each of the first main blocks;
means for dividing the image into a number of second main blocks using at least one second block raster, a position of the at least one second block raster being shifted relative to a position of the first block raster, and means for performing a second estimation process to provide a motion vector for each of the second main blocks;
means for forming a number of subblocks, wherein each of the subblocks is formed by an intersection region between one of the first main blocks and at least one of the second main blocks; and
means for generating a motion vector for each subblock as a function of the estimated motion vectors of the first and second main blocks whose intersection forms that subblock, each of the generated motion vectors being coordinated with the particular subblock.
15. A method for determining motion vectors that are coordinated with individual image regions of an image, comprising:
dividing the image into a number of first main blocks using a first block raster and performing a first estimation process to provide a motion vector for each of the first main blocks;
dividing the image into a number of second main blocks using at least one second block raster, a position of the at least one second block raster being shifted relative to a position of the first block raster, and performing a second estimation process to provide a motion vector for each of the second main blocks;
dividing the image into a number of second main blocks using at least one second block raster, a position of the at least one second block raster being shifted relative to a position of the first block raster, and performing a second estimation process to provide a motion vector for each of the second main blocks;
dividing the image into a number of third main blocks using at least one third block raster, a position of the at least one third block raster being shifted relative to the position of the first block raster and further being shifted relative the position of the second block raster, and performing a third estimation process to provide a motion vector for each of the third main blocks;
forming a number of subblocks, each of the subblocks being formed by an intersection, region between one of the first main blocks, at least one of the second main blocks, and at least one of the third main blocks; and
generating a motion vector for each subblock as a function of the estimated motion vectors of the first, second, and third main blocks whose intersection forms that subblock, each of the generated motion vectors for the subblocks being coordinated with the particular subblock;
where the dividing the image into the number of first main blocks, the dividing the image into the number of second main blocks, the dividing the image into the number of third main blocks, the forming a number of subblocks, and the generating the motion vector are performed by a video image processing device.