一种快速自适应RSUSAN角点检测算法

根据图像边缘灰度的渐变性，我们重新定义SUSAN(Small Univalue Segment Assimilating Nucleus)算法中小核值相似区；并找到一种更为有效和简便的计算小核值相似区面积的方法；在此基础上提出了RSUSAN(Redefined SUSAN)角点检测算法。与经典的角点检测算法SUSAN、MIC(Minimum Intensity Change)相比，RSUSAN具有角点检测准确性高，计算简单，运算速度大为提高等优点。对于模糊、噪声大的图像本文还进一步提出了采用自适应平滑和RSUSAN相结合的方法，称为自适应RSUSAN算法。实验证明，相比较SUSAN、MIC算法而言，自适应RSUSAN算法没有显著地增加计算量，而且在对模糊、噪声大的图像进行角点检测时，虚报及漏检概率大大减少，对噪声的鲁棒性好，角点检测位置精确。

Rapid Adaptive RSUSAN Algorithm of Corner Detection

Based on the gradual change of the image edge gray .a new redefinition of the region of the SUSAN (Small Univalue Segment Assimilating Nucleus )and a more efficient and simple method to compute it are suggested. And thus a new corner detection algorithm RSUSAN is proposed. The experiment shows that, compared with SUSAN and MIC.such algorithm is of simpler process,more rapid speed and more high veracity. In addition,in order to deal with the blurred and noisy image ,the way of combination of adaptive smoothing and such new algorithm is put forward,and the result of applying this adaptive RSUSAN corner detection algorithm to blurred and noisy image shows that it has similar numeration amount,but has better effect than SUSAN and MIC.