IMAGE KEY POINTS DETECTION AND DESCRIPTION BASED ON ESTIMATES OF PARTIAL DERIVATIVES

This paper considers a detecting and describing image keypoints method based on estimates of partial derivatives of the first and second orders. Vector masks are used to obtain estimates of partial derivatives in the vicinity of a keypoint. A brief description of known methods using an approach based on estimates of partial derivatives is given. An original method adapted for operation in real-time computer vision systems is proposed. The results of experiments are presented and prospects for further development are outlined.