摘要
提出了一种基于USAN的改进的角点检测算法。该算法在原有SUSAN算法的基础上做了如下改进:使用一个3×3的方形预检测窗口对图像的像素进行预检测,在精确检测角点前剔除掉大部分的背景点、边界点及脉冲噪声点,提高了算法的效率;根据图像不同区域对比度不同的特性,采用根据对比度自动调节核心点与其邻域像素的灰度差值门限的方法,使所检测出的角点分布均匀;利用基于USAN定义的角点所应具有的特征(角的边缘及USAN的连续性)来剔除伪角点,降低了角点虚报和漏检的发生率。仿真实验证明了该文所提出的算法具有抗噪能力强、运算量小的特点,适于实时实现。
To avoid the disadvantages of SUSAN. a new algorithm based on USAN is proposed. Comer pre-detection, which utilizes a window of three multiply three for detecting the candidate corners, can reject background, edges and noisy pixels. Intensity discrepancy between nucleus is redefined and its neighbors based on different area have different contrast. This makes the corners well-distributed. False corners are removed according to corner features, which is defined based on USAN. Experimental results show this comer detection method has good capabilities of detection and localization in different contrast image.
出处
《计算机工程》
CAS
CSCD
北大核心
2007年第22期232-234,共3页
Computer Engineering
基金
黑龙江省教育厅科学基金资助项目(10551115)
北京印刷学院院选人才引进基金资助项目
