2021/04/10 by Mansi Sharma, Sharma, Mansi, Santosh Kumar +1
Computer Science · #Advanced Vision and Imaging #Video Coding and Compression Technologies #Advanced Image Processing Techniques
paper · pdf · doi:10.48550/arxiv.2104.04678
The compression quality losses of depth sequences determine quality of view\nsynthesis in free-viewpoint video. The depth map intra prediction in 3D\nextensions of the HEVC applies intra modes with auxiliary depth modeling modes\n(DMMs) to better preserve depth edges and handle motion discontinuities.\nAlthough such modes enable high efficiency compression, but at the cost of very\nhigh encoding complexity. Skipping conventional intra coding modes and DMMs in\ndepth coding limits practical applicability of the HEVC for 3D display\napplications. In this paper, we introduce a novel low-complexity scheme for\ndepth video compression based on low-rank tensor decomposition and HEVC intra\ncoding. The proposed scheme leverages spatial and temporal redundancy by\ncompactly representing the depth sequence as a high-order tensor. Tensor\nfactorization into a set of factor matrices following CANDECOMP PARAFAC (CP)\ndecomposition via alternating least squares give a low-rank approximation of\nthe scene geometry. Further, compression of factor matrices with HEVC intra\nprediction support arbitrary target accuracy by flexible adjustment of bitrate,\nvarying tensor decomposition ranks and quantization parameters. The results\ndemonstrate proposed approach achieves significant rate gains by efficiently\ncompressing depth planes in low-rank approximated representation. The proposed\nalgorithm is applied to encode depth maps of benchmark Ballet and Breakdancing\nsequences. The decoded depth sequences are used for view synthesis in a\nmulti-view video system, maintaining appropriate rendering quality.\n