关键词: 视差梯度, 聚类, 相似测度

Abstract:

This paper proposes a method to extract corners from the image by the Harris algorithm, and gets rough matched corner set by similar measures. And then compares the simple clustering method and the disparity gradient constraint method for prepurring matching corners. The experimental results show that the clustering method is superior to the parallax Gradient constraint method. It also needs to further improve the efficiency of RANSAC algorithm in order to make it suitable for realtime processing of the occasion.

Key words: disparity gradient, clustering, similarity measure