A Prediction Mode Based on Sub-Block Position Swapping of Predicted Block for Video Coding

Due to the increasing demand of screen content video in existing video applications, newer video coding standards such as HEVC, AVS3, VVC, etc. are designed with a set of screen content coding tools targeted for them. Intra block copy mode, as a core tool for screen content coding in VVC standard, its coding performance is mainly affected by the similarity between the predicted block and the coding unit and the bit consumption of side information such as block vector. In this paper, we propose an intra block copy mode based on the sub-block position swapping of predicted blocks. The proposed algorithm is focused on reducing the consumption of side information and improving the compression performance by fetching more similar predicted blocks. It is a further optimization of the intra block copy technology so that it can handle some common text-heavy screen content video more efficiently. The proposed prediction mode is implemented on top of one of latest VVC reference software for performance evaluation and the experimental results show that the algorithm achieves 0.28% BD-rate gain in AI configuration.