Joint Data Partition and Rate-Distortion Optimized Mode Selection for H.264 Error-Resilient Coding

Data partitioning (DP) is an efficient error-resilient video coding tool. Its contribution to performance improvement in the error-prone environment arises from the superior error concealment mechanisms that are available with the help of protected data partitions. Since error-concealment in terms of DP is closely related to coding mode, it is desirable to have an optimized coding mode selection scheme. However, the existing coding mode selection techniques usually assume that the same error-concealment mechanism is used for a block when it is lost, and the associated distortion also remains the same. Obviously, this assumption is not true when DP involves. In this paper, a generalized end-to-end distortion model is proposed for the rate-distortion optimized coding mode selection, which fully utilizes the superior error-concealment mechanism in terms of DP. The proposed distortion model is also advantageous in the suppression of approximation errors caused by pixel average operations such as sub-pixel interpolation and deblocking filter. Therefore, it can lead to a low-complexity solution for real-time applications such as live streaming