Abstract
This paper proposes a new algorithm for error concealment in the presence of packet loss for block based video CODECs. It uses a one-pixel overlap block structure. Without any smoothness assumption on block boundaries, this method provides consistently high performance in lost block recovery. The one-pixel overlap generally decreases the compression ratio for a fixed quantization step size. However, increases in PSNR obtained by averaging overlapping pixels largely compensates the effect. Experimental results show that there is almost no performance decrease in PSNR vs. compression ratio for the `Football' and `Miss America' videos when using 9×9 blocks with one-pixel overlap (in place of the usual 8×8 blocks with no overlap).
| Original language | English (US) |
|---|---|
| Pages | 105-108 |
| Number of pages | 4 |
| State | Published - 1996 |
| Event | Proceedings of the 1996 IEEE International Conference on Image Processing, ICIP'96. Part 2 (of 3) - Lausanne, Switz Duration: Sep 16 1996 → Sep 19 1996 |
Other
| Other | Proceedings of the 1996 IEEE International Conference on Image Processing, ICIP'96. Part 2 (of 3) |
|---|---|
| City | Lausanne, Switz |
| Period | 9/16/96 → 9/19/96 |
ASJC Scopus subject areas
- Hardware and Architecture
- Computer Vision and Pattern Recognition
- Electrical and Electronic Engineering