Hierarchically layered adaptive median motion vector generation and smoothing

Abstract

Methods and systems for hierarchically layered adaptive median motion vector smoothing are disclosed. Aspects of one method may include generating motion vectors (MVs) for video pictures at each level of a hierarchical motion estimation process, where each level may use different resolution video pictures. The MVs may be smooth filtered at each level to reduce or remove spurious MVs. The smooth filtering may comprise scalar median filtering and/or vector median filtering. The smooth filtering may receive as inputs a plurality of MVs, for example, the MV being filtered and eight MVs from the surrounding video blocks. The vector costs of the nine MVs may be compared to a threshold vector cost, and those MVs with costs above the threshold value may be discarded. The threshold value may be dynamically changed based on, for example, content of the video pictures and/or to adjust a window size of the smoothing filter.

1 . A method, comprising:
 receiving a picture at a first resolution;   subsampling the picture to provide a subsampled picture at a second resolution;   generating a first motion vector for a second resolution block of the subsampled picture relative to a second resolution block of another picture;   smooth filtering the first motion vector;   scaling the smooth filtered, first motion vector; and   generating a second motion vector for a first resolution block of the picture relative to a first resolution block of the another picture, the generating of the second motion vector based on the scaled, smooth filtered first motion vector.   
     
     
         2 . The method of  claim 1 , further comprising subsampling an input picture to provide the picture at the first resolution. 
     
     
         3 . The method of  claim 2 , further comprising smooth filtering the second motion vector. 
     
     
         4 . The method of  claim 3 , further comprising scaling the smooth filtered, second motion vector. 
     
     
         5 . The method of  claim 4 , further comprising generating a third motion vector for an input picture resolution block of the input picture relative to an input picture resolution block of the another picture, the generating of the third motion vector based on the scaled, smooth filtered second motion vector. 
     
     
         6 . The method of  claim 5 , further comprising smooth filtering the third motion vector and outputting the smooth filtered, third motion vector. 
     
     
         7 . The method of  claim 2 , wherein the subsampling of the input picture and the subsampling of the picture are implemented at a same subsampling rate. 
     
     
         8 . The method of  claim 2 , wherein the subsampling of the input picture and the subsampling of the picture are implemented at a different subsampling rate. 
     
     
         9 . The method of  claim 1 , wherein the smooth filtering comprises vector cost thresholding and median filtering. 
     
     
         10 . The method of  claim 9 , wherein the vector cost thresholding comprises:
 receiving the first motion vector; and   receiving motion vectors adjacent the first motion vector, wherein the median filtering includes as input the first motion vector and any of the adjacent motion vectors that have a cost threshold at or below a threshold vector cost   
     
     
         11 . The method of  claim 10 , wherein the median filtering comprises vector median filtering or scalar median filtering. 
     
     
         12 . The method of  claim 10 , wherein the vector cost threshold is based on content of the picture. 
     
     
         13 . The method of  claim 10 , further comprising adaptively adjusting a size of a median filter window. 
     
     
         14 . A system, comprising:
 a processor comprising a search engine, the search engine configured to perform multi-layer hierarchical motion estimation, wherein for N levels and a received input picture, the search engine is configured to:
 generate N-1 lower resolution pictures for use by respective N-1 levels, each of the N-1 levels comprising a different picture resolution; 
 generate motion vectors for each of the N levels of motion vector generation; 
 smooth filter at least one of the motion vectors for each of the N levels; and 
 scale the smooth filtered motion vectors of the N-1 levels, wherein the generation of motion vectors at each of the N levels is based on a present and prior picture at the respective level resolution and based on the smooth filtered and scaled motion vectors generated at an adjacent lower level. 
   
     
     
         15 . The system of  claim 14 , wherein the search engine is configured to generate N-1 lower resolution pictures by subsampling, the subsampling at plural levels implemented at a same subsampling rate. 
     
     
         16 . The system of  claim 14 , wherein the search engine is configured to generate N-1 lower resolution pictures by subsampling, the subsampling at plural levels implemented at a different subsampling rate. 
     
     
         17 . The system of  claim 14 , wherein the search engine is configured to smooth filter using vector cost thresholding and either vector median filtering or scalar median filtering. 
     
     
         18 . The system of  claim 17 , wherein the vector cost thresholding comprises:
 receiving the at least one of the motion vectors; and   receiving motion vectors adjacent the at least one of the motion vectors, wherein the median filtering includes as input the at least one of the motion vectors and any of the adjacent motion vectors that have a cost threshold at or below a threshold vector cost   
     
     
         19 . The system of  claim 18 , wherein the vector cost threshold is based on content of the picture. 
     
     
         20 . A non-transitory, computer readable storage medium comprising executable code executed by a machine to cause the machine to:
 generate N-1 lower resolution pictures based on an input picture, the generated N-1 lower resolution pictures for use by respective N-1 levels, each of the N-1 levels comprising a different picture resolution;   generate motion vectors for each of the N levels of motion vector generation;   smooth filter at least one of the motion vectors for each of the N levels; and   scale the smooth filtered motion vectors of the N-1 levels, wherein the generation of motion vectors at each of the N levels is based on a present and prior picture at the respective level resolution and based on the smooth filtered and scaled motion vectors generated at an adjacent lower level.