The influence of a lossy compression on the quality of estimated depth maps

The paper considers the influence of a lossy compression of multiview video on the quality of estimated depth maps. We have encoded a set of views with commonly known compression techniques used in compact IP cameras: HEVC (MPEG H part 2 / H.265), A VC (MPEG 4 part 10 / H.264) and MPEG 2 part 2 / H.262. Basing on the decoded data we have estimated depth maps and rendered a set of the intermediate virtual views. A drop of the quality was studied by comparison of virtual views rendered out of the depth maps estimated from an uncompressed video. The obtained results show that with a reasonable compression rate, lower quality of virtual views is acceptable for real implementable FTV systems.

[1]  Marek Domanski,et al.  View Synthesis for Multiview Video Transmission , 2009, IPCV.

[2]  Krzysztof Wegner,et al.  Video and depth bitrate allocation in multiview compression , 2014, IWSSIP 2014 Proceedings.

[3]  Krzysztof Wegner,et al.  Fast depth estimation on mobile platforms and FPGA devices , 2015, 2015 3DTV-Conference: The True Vision - Capture, Transmission and Display of 3D Video (3DTV-CON).

[4]  Krzysztof Wegner,et al.  Fast view synthesis using platelet-based depth representation , 2014, IWSSIP 2014 Proceedings.

[5]  Antonio Ortega,et al.  On Dependent Bit Allocation for Multiview Image Coding With Depth-Image-Based Rendering , 2011, IEEE Transactions on Image Processing.

[6]  Krzysztof Wegner,et al.  A practical approach to acquisition and processing of free viewpoint video , 2015, 2015 Picture Coding Symposium (PCS).

[7]  Yannick Morvan,et al.  Joint depth/texture bit-allocation for multi-view video compression , 2010 .

[8]  Toshiaki Fujii,et al.  FTV for 3-D Spatial Communication , 2012, Proceedings of the IEEE.

[9]  Thomas Wiegand,et al.  3-D Video Representation Using Depth Maps , 2011, Proceedings of the IEEE.

[10]  Krzysztof Wegner,et al.  Quantization optimization in multiview plus depth video coding , 2014, 2014 IEEE International Conference on Image Processing (ICIP).