Real-time HDR video tone mapping for high efficiency video coding

Abstract

HDR videos become more common by advances in the display technology but require additional computation of tone mapping for handling the difference of the device dynamic ranges. However, the key challenge of HDR video tone mapping is to calculate luminance adaptation with both spatial and temporal variation for every pixel, resulting in heavy computational cost. It has limited the applicability of HDR video on various devices. In this work, we address the challenges of computational cost in HDR video tone mapping. Since video streams are delivered via the high efficiency video coding standard, we propose a novel video tone-mapping method that takes advantage of precomputed information in the coding stage of video compression. The video compression standard calculates coding blocks of different block sizes and their motion vectors to reduce data redundancy of a video stream. We are inspired to utilize these precomputed coding blocks and motion vectors in the decoding stage, to achieve both local adaptation and temporal coherence of luminance for video tone mapping. Our real-time video tone-mapping method can reduce computational costs significantly, while the quality is comparative to state-of-the-art video tone-mapping methods.