Error resilience video coding parameters and mechanisms selection with End-to-End rate-distortion analysis at frame level

Abstract

To improve the quality of video transmission, a fast error resilience coding method based on frame level rate-distortion analysis is proposed. To constrain the accumulated error propagation distortion and error concealment distortion simultaneously, reference frame selection and intra/inter mode decision are jointly used with redundant pictures. An adaptive multiple redundant picture (AMRP) coding mechanism is used for redundant picture coding with the adaptively estimated number of redundant pictures and encoding parameter for each specific redundant picture. The encoding parameters of different frames are adjusted based on the distortion propagation. We propose a statistical model for efficiently estimating the distortion and rate of the primary and the redundant picture. The total distortion and rate of the primary and the redundant picture are then formulated as a function of the quantization parameter, the temporal prediction distance, and the error resilient configuration. Lastly, the end-to-end rate-distortion optimized selection of the encoding parameters and coding structure is efficiently performed considering error propagation. Experimental results demonstrate that the proposed algorithm exhibits significant performance gains over the state-of-the-art error-resilient encoding methods.