Aiming at problem of low matching efficiency and large error of dense stereo matching algorithm based on traditional region growing

an improved dense stereo matching algorithm is presented based on the technique of region growing.Firstly

BRISK feature points are considered to be the seed points of region growing

and texture images are used to reduce the impact of varying illumination in local area in different perspectives views

then the constraint criterion of direction and epipolar line distance is added to narrow the range of the adaptive searching window.Experiment demonstrates that the algorithm has higher matching precision and wider matching range than traditional one.