Dual region based color channel registration for multiple color filtered aperture image

Abstract

This paper presents a dual region based cross-channel registration for the dual color-filter aperture (DCA) image. Despite a DCA image contains depth information, disparity calculation between its color channels is a complex problem. The proposed registration method consists of four steps: (1) dual region division, (2) disparity map generation using Cross-Channel Normalized Gradient (CCNG), (3) disparity refinement using simple linear iterative clustering(SLIC) superpixel and (4) reconstruction aligned color channel using inpainting. Experiment result shows that the proposed method performs significant for DCA image registration and depth estimation.