Complexity and implementation analysis of synthesized view distortion estimation architecture in 3D High Efficiency Video Coding

Depth-based 3D formats for state of the art High Efficiency Video Coding standard facilitate the synthesis of virtual views with a brief binary representation. Depth maps coding error results in synthesis artefacts for virtual views rendered during Depth Image Based Rendering (DIBR) process. Based on depth maps distortion, synthesized view distortion estimation models are integrated in 3D-High Efficiency Video Coding (3D-HEVC) to cater the synthesis artefacts. This paper presents complexity and implementation analysis of synthesized view distortion estimation model integrated in Emerging 3D-HEVC standard. Profiling of the reference software of 3D-HEVC is performed and computational hotspots are identified based on the profiling information. Profiling results show that the Renderer model integrated in 3D-HEVC, for Rate Distortion Optimization (RDO) of depth maps coding by estimation of synthesized view distortion, takes about (18-26%) of total encoding time and is identified as one of the computational complex part of the 3D-HEVC standard. Algorithmic and implementation complexity analysis of the renderer model is presented in this paper.