Abstarct: Disparity is the diffrence position between two corresponding points from left and right images which is consider one of those images refrences. Non_existent of coresponding point for each pixei from reference image against other image should be recognize in right way. The latest proposed algorithms with accuracy to extract a map in sub pixel scale from texture less regions, repeated texture regions, beside depth discontinuity regions, occluded regions, horizontal and vertical slant regions and also in cylindrical and spherical surfaces are far from ideal. Our purpose in this article is considering new idea and gather them for optimizing disparity map that include each above items. From the varity of stereo vision algorithms like pixel_baxsed, feature_baxsed and area_baxsed methods, we used area_baxsed algorithms in this job. Algorithm program include 3 step. The first step is pre_processing which we used average filter. The second step we used adaptive weight windows around selected pixel in CIE lab and I1I2I3 color space. Also to reduce the search space from epipolar and rang of disparity constraints is used. The performance of this step is a premium disparity map. The last step is post_processing which we used a mwdian weight filter as guided filter. According to the result, suggested algorithm from boundury regions in correct way and also we have better result in other job in chapter 5 that is compared with them.