Image can be presented in graph or grid graph. Segmentation is widely used in computer vision. Image segmentation is done in various approaches. One approach to image segmentation is the use of a minimum spanning tree. Almost every method of determining disparity in stereo vision uses image segmentation. The method in this paper is in accordance with the method of Segment-Based Stereo Matching using belief propagation and self-adapting dissimilarity measure with improvements for segmentation, and apply local window-based matching steps. The disparity calculation method built with minimum spanning tree approach is one alternative that can be used to calculate disparity for image pair with simple stereo configuration.