Name
Abstract | ||
|---|---|---|
Whole-frame loss of the compressed video is very common in transmission over error-prone networks since each coded picture is usually packetized into one single packet in order to reduce the bitstream overhead for transmission. In this paper we present an adaptive frame recovery algorithm which innovatively introduces the three-dimensional recursive search (3DRS) motion estimation method into fram reovr (FR), and dynamically selects between 3DRS based recovery and motion vector copy (MVC) based recovery according to the statistics of motion activity of previous frames. If the motion activity of the frame is large, we adopt 3DRS-based frame recovery. Otherwise MVC-based FR is used. For the former method, we first perform the modfief 3DRS to re-estimate the motion vectors (MVs) of the previous frame considering that the available motion information of previous frames is not close to the true motion trajectory. Then the MVs are extrapolated and refined as the MVs of the lost frame. The missing frame is recovered using motion compensation. For MVC-based FR, motion information of previous frames, derived from the decoder is reused. Experimental results show that our proposed solutions can achieve significant improvements in both PSNR and visual quality. |
Year | DOI | Venue |
|---|---|---|
2007 | 10.1109/SIPS.2007.4387633 | SiPS |
Keywords | Field | DocType |
frame recovery,3drs motion estimation,mv copy,mv extrapolation,error concealment,motion compensation,decoding,psnr,statistics,nonvolatile memory,three dimensional,video compression,motion estimation | Reference frame,Computer vision,Block-matching algorithm,Quarter-pixel motion,Computer science,Motion compensation,Residual frame,Artificial intelligence,Inter frame,Motion estimation,Motion vector | Conference |
ISSN | ISBN | Citations |
1520-6130 E-ISBN : 978-1-4244-1222-8 | 978-1-4244-1222-8 | 0 |
PageRank | References | Authors |
0.34 | 5 | 5 |