Optimizing ZNCC calculation in binocular stereo matching

Stereo matching is a crucial and challenging step in binocular vision measurement. Robust zero-mean normalized cross-correlation (ZNCC) is widely used for stereo matching. However, direct calculation by using the prevalent ZNCC algorithm is computationally expensive because of the large number of redundancies that directly affect execution time. Therefore, this study proposes a fast method for the reliable computation of the similarity measure through ZNCC for stereo matching. We divide the standard ZNCC function into four independent parts, and this can efficiently reduce computational complexity. Furthermore, a storage strategy is proposed to store calculation results by column and apply a circular queue to the entire matching process. The rapid calculation of the template relies on the position of the pixels in the given image, which is based on the relevant characteristics of adjacent pixels. Invoking the stored calculation template value can help significantly reduce computational complexity. The proposed algorithm was tested on a 2.6-GHz computer with different sizes of images from the Middlebury Stereo Datasets, and the results reveal a remarkably shorter execution time.