An error concealment adaptive framework for intra-frames

The performance of error concealment (EC) for damaged intra-frames (or I-frames) is very important for real time video transmission via error-prone network. Conventional algorithms essentially exploit either copying or interpolation from spatial neighbors and suffer from poor visual quality. This paper proposes an intra-frame EC adaptive algorithm framework (I-ECAF) which gives priority to temporal EC as the first, spatial copying as the second and spatial interpolation as the last. I-ECAF evaluates their feasibility by measuring the temporal or spatial similarity per a recent image quality metric, i.e., the weighted peak signal-to-noise ratio (WPSNR). Demonstrated by extensive experiments, the proposed I-ECAF gains attractive concealment accuracy, an increase from 0.06dB to 1.14dB of PSNR in comparison with the algorithm embedded within JM18.0.