Distributing candidate vectors based on local motion complexity
A system and method for distributing candidate motion vectors divides a picture frame ( 110 ) into a plurality of segments ( 120 ), where each segment ( 120 ) includes a plurality of pixel blocks ( 130 ). The local motion complexity of each segment may then be measured by determining a Sum-of-Absolute Differences (SAD) between pixel blocks ( 130 ) of the frame ( 110 a ) and pixel blocks ( 130 )of an adacent frame ( 110 b ) and summing the SAD values for the pixel blocks ( 130 ) within each segment ( 120 ). The maximum, minimum and average SAD value for the segments may be used to define a distribution function that distributes candidate vectors to conform with a desired maximum, minimum and average number of candidate vectors per block. This distribution function is then used to determine the number of candidate vectors to assign to pixel blocks ( 130 ) based on the measured SAD value of the associated segment ( 120 ).
1. A method for distributing candidate motion vectors, the method comprising:
dividing a picture frame into a plurality of segments using a processor, each segment comprising a plurality of pixel blocks;
measuring local motion complexity for each segment using the processor; and
assigning a number of candidate motion vectors to pixel blocks within each segment based on the measured local motion complexity using the processor, wherein the number of candidate motion vectors assigned to pixel blocks within one of the segments is different from the number of candidate motion vectors assigned to pixels blocks within another one of the segments,
wherein the measuring local motion complexity further comprises:
determining a sum-of-absolute differences between pixel blocks of the picture frame and corresponding pixel blocks of an adjacent frame; and
summing the measured sum-of-absolute differences associated with of pixel blocks within each segment,
wherein the assigning further comprises using a distribution function configured to assign the number of candidate vectors based on the measured local motion complexity of each segment, and
wherein the distribution function is based on a maximum, minimum or average of the measured sum-of-absolute differences (SAD) of the segments using the following respective relationships:
N _max= a +b *SAD_max + c *SAD_max*SAD 13 max,
N _min= a +b *SAD_min + c *SAD_min*SAD_min, and
N _av= a +b * SAD_av + c *SAD_av*SAD_av.
2. The method of claim 1 , wherein the distribution function is further based on predetermined values for a maximum, minimum or average number of candidate vectors per block.
3. The method of claim 1 , further comprising performing motion estimation on the pixel blocks using the number of candidate vectors assigned to each pixel block.
4. A system for distributing candidate vectors, the system comprising a processor configured for:
dividing a picture frame into a plurality of segments, each segment comprising a plurality of pixel blocks;
measuring local motion complexity for each segment; and
assigning a number of candidate motion vectors to pixel blocks within each segment based on the measured local motion complexity, wherein the number of candidate motion vectors assigned to pixel blocks within one of the segments is different from the number of candidate motion vectors assigned to pixels blocks within another one of the segments,
wherein the measuring local motion complexity comprises:
determining a sum-of-absolute differences between pixel blocks of the picture frame and corresponding pixel blocks of an adjacent frame; and
summing the measured sum-of-absolute differences associated with of pixel blocks within each segment,
wherein assigning comprises using a distribution function configured to assign the number of candidate vectors based on the measured local motion complexity of each segment, and
wherein the distribution function is based on a maximum, minimum or average of the measured sum-of-absolute differences (SAD) of the segments using the following respective relationships:
N _max= a +b *SAD_max + c *SAD_max*SAD_max,
N _min= a +b *SAD_min + c *SAD_min*SAD_min, and
N _av= a +b * SAD_av + c *SAD_av*SAD_av.
5. The system of claim 4 , wherein the distribution function is further based on predetermined values for a maximum, minimum, or average number of candidate vectors per block.
6. The system of claim 4 , wherein the processor is further configured for performing motion estimation on the pixel blocks using the number of candidate vectors assigned to each pixel block.