This Motion vectors of depth-maps in multiview and free-viewpoint videos exhibit strongspatial as well as inter-componentclustering tendency. In thispaper, we present a schemethat encodes the motion vector information at the frame level to efficiently exploit thesecorrelations. It arranges the motion vectors ofall the macroblocks of a frame into a 3Dvolume by considering that the horizontal (x-axis) and vertical (y-axis) displacementcomponents belong to adjacent planes (p-axis). Information about the positions ofmacroblocks of identical coding mode is then extracted into different bitmaps. Theproposed scheme first compresses the multidimensional bitmaps of macroblock modeinformation and then encodes only thenon-zero components of motion vectors.
|Number of pages||1|
|Publication status||Published - 2015|
|Event||Data Compression Conference - Snowbird, Utah, USA, New Zealand|
Duration: 07 Apr 2015 → 09 Apr 2015
http://www.cs.brandeis.edu/~dcc/Program.html (conference programs)
|Conference||Data Compression Conference|
|Period||07/04/15 → 09/04/15|