| 摘要 |
In real-time video applications, where unreliable networks are commonplace, corrupted video packets can affect adversely the visual quality. In this patent application, we present an improved method and system for video error correction based on maximum likelihood. Rather than discarding corrupted video packets, the method estimates the likeliest syntactically valid video slice content based on these packets. We present two embodiments, which permits solving the problem at the slice-level, and a simplified solution operating at the syntax element-level. The performance of the method is evaluated using the H.264 baseline profile. Unlike existing video error concealment methods, we correct errors in the bitstream instead of reconstructing missing pixels. Simulation results show that the proposed method yields improved visual quality, and is also computationally simpler than existing state-of-the-art error concealment methods. A corresponding system for video error correction is also provided. |
