The invention discloses a stereoscopic parallax optimizing method based on an anti-texture cross and a weight cross. The problem that a disparity map obtained through a disparity optimizing method in the prior art is inaccurate is mainly solved. The stereoscopic parallax optimizing method includes the steps that 1), mistaken point detection is carried out on the original horizontal parallax disparity map to obtain a mistaken point set; 2), the original color image is preprocessed to extract the structure information; 3), the structure information is used for setting up the support region for each mistaken point; 4), the parallax price is calculated for each point in the support region; 5), by using the support region, the weight cross is set up through calculating weights of adjacent points; 6), for each layer of selectable parallax, previous-next information transmission parallax price aggregation is carried out on the weight cross; 7), the optimal parallax of the mistaken points is selected through parallax aggregation gross information content; 8), the step from 3) to 7) are repeated until all the mistaken points are updated. Parallax optimization can be fast and accurately carried out on the original parallax disparity map, and the stereoscopic parallax optimizing method can be used for binocular stereo matching.