Abstract: In order to solve the problem of low calculation speed of the Binary Robust Invariant Scalable Keypoints (BRISK) algorithm and extreme density of the feature points extracted, an improved BRISK feature extraction algorithm based on quadtree (Quad-BRISK algorithm) with the method of quadtree homogenization is proposed. After the feature points with scale invariance are extracted and detected on the generated image pyramid, the feature points are divided with the quadtree method, and then the direction of the feature points and the BRISK descriptor are calculated. After coarse matching, filtering and purification, the precise matching images are finally obtained. Using the feature comparison test dataset of Mikolajczyk and Schmid, the SIFT, ORB, BRISK and Quad-BRISK algorithms are tested and compared. The experimental results show that the Quad-BRISK algorithm can not only extract more stable feature points but also improve the matching accuracy and calculation speed of feature points.