Improved Stereo Matching Algorithm Using Contrast Limited Adaptive Histogram Equalization

Abstract

Stereo matching is one of the most active areas in the field of computer vision. The main goal of matching is to determine disparities i.e., difference in locating corresponding pixels and the recovery of the 3D structure of a scene. The recovery of an accurate disparity map still remains challenging task. In this work, we present a novel AHE based feature point stereo matching algorithm with global energy minimization technique. The initial disparity map is estimated by considering matching SURF key points between two images inside each homogeneous color region using clustering technique which is done by grouping the similar pixels in an image which gives the image disparity. Then RANSAC based plane fitting technique is applied to find the depth map which relies on accuracy of the pixel disparities inside the homogeneous color regions. Finally, the disparity map is further enhanced by incorporating energy constraints on smoothness between neighboring regions using median average filter and contrast limited adaptive histogram equalization. This above modified methodology is tested on Middlebury test bed which indicates that this method gives better performance with the current state-of-the-art stereo matching algorithms with reduced computational effort in matching process.
Copyright © 2013 Praise Worthy Prize - All rights reserved.