Hash image transform for accelerated screen content encoding in HEVC

In this paper, we present a hash-based framework for acceleration of a block-based hybrid video encoder. The framework consists of a hash-based transform, a hash-based motion analysis, and a finial collaboration of the analysis information for different block sizes. The hash-based transform generates an image where each sample represents a complete block. Thus, in the hash-based motion analysis, only a comparison of samples from the hash images is performed instead of complete blocks of the actual images. The results from the motion analysis are collaborated for different block sizes, which is then used to guide a block-based video encoder. The proposed framework was implemented on top of HM-16.7SCM-6.0 and has been evaluated for encoding of screen content video signals. Simulation results show that the proposed framework achieves significant runtime reductions of up to 82.8% while approximately keeping or lowering the bitrate by up to 6.1% compared to the HEVC reference software HM-16.7_SCM-6.0 for low-delay-P-frame coding.