An adaptive corner detecting for real-time applications

We propose an adaptive and fast corner detector based on a double-circle mask to meet real-time applications. First, we calculate the variance of the nucleus of every pixel in the mask. Second, we present an adaptive threshold of a flat area to detect whether the nucleus is on a flat area to avoid the furthermore detecting. Then a corner response function is used to detect the corner candidates. Finally, we propose a false corner removal algorithm to ensure the quality of the corners. The experimental results show that the proposed method performs better in time consuming and corner detecting, and also is robust to noises.