Refined side information algorithm based on motion features of video sequences
Zhu Jin-xiu · Jisuanji gongcheng yu sheji · 2012
To solve the problem that the generated error of areas doing nonlinear movement is greater than other regions in side information(SI),a method based on motion features to refine the SI is proposed.First in the decoding side,initial SI is produced by motion compensated temporal interpolation(MCTI),then linear and non-linear motion areas in SI are distinguished by the sum of the for-and back-ward motion vectors(MVs).After correcting the false MVs by motion vector filter(MVF),hierarchical motion estimation(HME) is utilized for the further processing of blocks belonging to non-linear area to obtain more accurate motion information.The quality comparison is made between new SI frames and the initial ones,which finally indicate that the average PSNR of new ones improve 1dB.