Edge detection is a basic problem of image processing and one of the hot areas of research. Today' s mainstream edge detection operator theory and methods still have some shortcomings,such as the poor balance between positioning accuracy and noise filtering. The traditional Canny operator is lack of the automation process capability,aiming at breaking the limit,an efficient method of adaptive Canny algorithm is proposed. The method chooses the filter to suppress the noise adaptively according to image noise features. By expanding gra-dient calculation to 4-directions,non-maximal suppression is simplified to refine the edge,the adaptive threshold is computed by the maximum value,average and variance of gradient amplitude to extract edge. In order to estimate the detection capability of each operator, a comprehensive performance evaluation method based on edge characteristics is proposed,judging the detection results by comparing the external features,location accuracy,anti-noise performance and computing time. The theoretical and experimental result shows that the improved operator is easy to realize,which can position the edge accurately,with high detection precision,can process different noise im-age flexibly,fulfilling the requirement of practical application.%边缘检测是图像处理的基础问题之一,也是该领域的研究热点.而当今主流边缘检测算子的理论和方法尚存在一些不足,如定位精度与噪声滤除间的平衡性较差等.针对Canny算子自适应性较差等不足,提出一种改进方法.根据图像噪声特征自适应地选择滤波器抑制噪声,扩展梯度运算至4方向并简化边缘细化处理过程,根据梯度幅值的最大值、平均值和方差等计算自适应阈值提取边缘.为评价各算子的检测能力,基于边缘的特征提出综合性能评估方法,通过比较边缘的外部特征、定位精度、抗噪性和运算时长等判断边缘检测效果.通过分析实验数据发现,改进算子实现简单,能准确定位边缘,检测精度高,可灵活处理不同噪声图像,能较好地满足实际应用的需要.
展开▼
