Abstarct: The process of extracting depth information from multiple two-dimensional images taken of the same scene is known as stereo vision. Spatial displacement is the difference between corresponding pixels from the left and right images, by taking one of the images as reference. In the past few decades have witnessed the development of hundreds of different stereo vision algorithms, this classification made it difficult to compare the various approaches to this problem. In this thesis we provide an overview of everything that approaches to solving stereo vision problem. Because we provide a method to extract real-time disparity map, between local and global stereo algorithms, we use local methods. In tests carried out over recent years, the most important element of the local stereo algorithms, matching cost aggregation methods expressed. Our preferred method, first, a feature-baxsed local stereo algorithm using harris corner detection are presented and then with using proposed matching cost aggregation methods around the window of interest pixel, method is proposed to recover the sparse disparity map that accuracy is better than previous methods. Then provided a local stereo algorithm baxsed on pixel to recover a dense disparity map using gamma correction in RGB color space on the pair of input images in the preprocessing, bilateral filter in the postprocessing and using matching cost aggregation methods expressed in sparse methods that a mean error at the presented databaxse in the literature is %5.1 and speed is desired.