H19: Improved Lossless Depth Image Compression

Fischer,R., Dittmann.P, Schroeder.C, Zachmann,G.

Abstract:
Since RGB-D sensors became massively popular and are used in a wide range of applications, depth data compression became an important research topic. Live-streaming of depth data requires quick compression and decompression. Accurate preservation of information is crucial in order to prevent geometric distortions. To fulfill these requirements and simultaneously achieve high compression ratios, custom algorithms are needed considering the unique characteristics of depth images. We propose a real-time capable lossless algorithm based on RVL, which achieves significantly higher compression ratios. The core elements are an adaptive span-wise intra-image prediction and parallelization. Additionally, we extended the algorithm by an inter-frame difference computation stage and evaluated the performance regarding different conditions. Lastly, the compression ratio can be further increased by a second encoder, circumventing the lower limit of four-bit per valid pixel of the original RVL algorithm. To do so, we did a lot of extensive experiments.