An improved Harris corner detection algorithm for low contrast image

摘要

In seas security supervision, we should conduct feature extraction and match for consecutive frames in order to effectively reduce the jitter caused by wind-induced vibration. Harris corner point detection algorithm is widely used in feature extraction for images. An improved algorithm for Harris corner detection is proposed in this paper since the feature points of images with large size, high pixel and low contrast ratio can not be extracted accurately, especially for seas security monitoring. To begin with, comparing with the eight pixel points around, we can select some possible corner points. And then, by curve fitting, we can choose the pixel points whose R values reach the maximum values in the X and Y directions as the another kind of possible corner points. On top of this, the above two kinds of possible conner points are processed by Susan operator so as to take out the peripheral points and pseudo points, and the best match points can be obtained. The experimental results on seas security applications show the effectiveness and feasibility of the proposed method.