Sparse disparity map computation from stereo-view images using segment based algorithm

In this paper, we propose a novel disparity map estimation method based on image segmentation and detection of feature points. Image segmentation and detection of feature points in images plays a very important role in stereo-view analysis. Firstly, K-Means based segmentation method is applied for segmenting the input images into regions. The aim of the segmentation is to simplify representation of an image into the form that is more suitable for analysis and further processing, yielding correct disparity estimates. Secondly, results of the image segmentation are used as an input of the SIFT-SAD algorithm to determine the disparity estimate. The proposed matching algorithm combines the Scale Invariant Feature Transform (SIFT) with the Sum of Absolute Difference (SAD). The obtained experimental results demonstrate that the performance of our method is competitive and the final disparity maps are close to the ground truth data.