Macropixel based fast motion estimation for plenoptic video compression

Abstract

A generalized plenoptic video coding method is proposed to realize fast motion estimation in integer pixel precision by exploiting the imaging structure correlation among macropixels. Based on the imaging structure correlation, collocated blocks in the neighboring macropixels are used to predict the current prediction unit (PU) and the 8-step neighborhood refinement improves the performance further. The experimental results demonstrate the proposed algorithm outperforms the state-of-the-art plenoptic compression methods, like Test Zone Search (TZS) algorithm embedded in the HEVC Test Model (HM), Rotation Scan Mapping (RSM) method and Light Field Multiview Video Coding algorithm (LF-MVC) with an average of 7.11%, 69.72% and 41.14% bitrate reduction, respectively, under the comparative computing complexity.