Morphology-based disparity estimation and rendering algorithm for light field images

Abstract

A light field camera can retrieve the depth information and reconstruct a stereoscopic image. The quality of light field image processing is highly dependent on whether the disparity of two adjacent micro images can be determined accurately. In this paper, we apply an algorithm based on dilating image segments to determine disparity precisely. With the dilation operation, the edge and corner information, which are important features for template matching, can be well preserved for each segment. Moreover, we also apply the techniques of superpixel-based object extraction and intensity compensation to further improve the performance. Simulations show that, with the proposed algorithm, a more accurate depth map and a rendering image with higher quality can be achieved.