Aiming at the problem of computational redundancy in cost aggregation in semi-global matching (SGM), a fast cost aggregation and disparity determination method based on the minimum cost path is proposed, which includes two stages: 1) Different from SGM, the cost aggregation is conducted by finding the path of minimum cost instead of updating the matching cost. Then the cost aggregation result is transformed from the aggregation cost to a set of disparity candidate values. The amount of cost aggregation calculations is no longer related to the disparity range. 2) After all the paths are aggregated and the disparity candidate value set is obtained, the reciprocal of the sum of the absolute difference between a single disparity candidate value and other disparity candidate values is used as its support. Counting the support of all disparity candidate values, the “winner takes all” strategy is employed to determine the final disparity value after the support of all disparity candidate values are counted. The experimental results of the dense matching of ZY-3 and GF-7 images verify the effectiveness of the method, which can increase cost aggregation processing efficiency by 90% while keeping the matching results consistent with SGM.