Title :
Motion vector recovery for H.264 based on mean shift
Author :
Zhan, Xuefeng ; Zhu, Xiuchang
Author_Institution :
Image Process. & Image Commun., Lab., Nanjing Univ. of Posts & Telecommun., Nanjing, China
Abstract :
In this paper, we describe a novel motion vector (MV) recovery method for H.264 which is based on mean shift procedure. In H.264, macroblock (MB) can be divided into several 4×4 blocks. Generally, MVs of these blocks in local region are very similar. To conceal the lost MB temporally, we consider the MV recovery of the lost MB as smooth operation in the neighboring area. The mean shift procedure is applied here to estimate MVs of lost blocks. Experimental results demonstrate their better performance.
Keywords :
error correction; video coding; H.264; error concealment; mean shift procedure; motion vector recovery; Clustering algorithms; Error correction; Image communication; Image processing; Interpolation; Kernel; Lagrangian functions; Motion estimation; Partitioning algorithms; Video compression; Error concealment; mean shift;
Conference_Titel :
Communications, 2009. APCC 2009. 15th Asia-Pacific Conference on
Conference_Location :
Shanghai
Print_ISBN :
978-1-4244-4784-8
Electronic_ISBN :
978-1-4244-4785-5
DOI :
10.1109/APCC.2009.5375642
