Adaptive Search Range of Template Matching for Plenoptic Video Coding under Microlens Constraints

This paper proposes a fast template matching (TM) based intra prediction for plenoptic video coding. While the conventional TM method has demonstrated itself as a potentially effective coding tool for plenoptic video, it suffers from a huge time complexity of TM search at both encoder and decoder. In this paper, we note correlation among matching pixels arising from the microlens array structure of plenoptic camera, and propose an adaptive search range (ASR) exploiting relationship between the two sizes of prediction unit and microlens. The experimental results show that our method can reduce encoding time by about 88% and 73% compared to the conventional TM method in the AI-Main and RA-Main conditions, respectively, with acceptable coding loss.