Improved NCC Stereo Matching Algorithm

Abstract

In the binocular stereo vision system, image matching is a key step to realize 3-D reconstruction. So many methods were proposed to solve the problem, among which the normalized cross correlation (NCC) algorithm is usually used because of its advantages such as high robust and precision, but its computing process is so complicated that difficult to apply on line. Therefore, the integral image and squared integral image are introduced, an improved NCC algorithm is proposed. The summation operation of pixels’ value in a certain region will be reduced greatly by using the algorithm, that is, it only requires to operate additions and subtractions of four pixels’ value. Simultaneously, mismatch areas in the reference image are eliminated to decrease the search range. The improved NCC algorithm not only guarantees the match quality but also improves the speed of execution greatly. Through the simulation experiment, the validity of it is verified.