Intelligent sub-patch texture synthesis algorithm for smart camera

Abstract

We propose an intelligent sub-patch texture synthesis algorithm for smart camera. Redundant human removal is a novel function in the next generation smart camera. In order to fill the lost region, texture synthesis is an important research topic. We present an efficient constrained texture synthesis algorithm which can past a patch each time and used in human removal. The proposed intelligent algorithm is suitable to be integrated in smart camera and used in image reconstruction. The searching neighborhood pixels can be changed due to different patch size. The intelligent algorithm consists of three steps: line-expanding, texture synthesis and block-smoothing. The boundary artifact can be removed by the block-smoothing method. The comparison of previous algorithms is also provided in this paper. We present a first algorithm used in constrained texture synthesis by regular synthesizing order.