图像边缘检测的分数阶微分算子研究

ON FRACTIONAL DIFFERENTIAL OPERATORS FOR IMAGE EDGE DETECTION

针对常用整数阶微分边缘检测算子不能较好保持图像纹理细节的不足,在4-方向的Roberts算子、Prewitt算子和Sobel算子的基础上利用0~1阶分数阶微分替换一阶微分,构造了3种用于图像边缘检测的0~1阶分数阶微分新算子。实验结果表明,所构造的3种分数阶微分算子不仅能有效地提取出图像的边缘信息,而且还能较大程度地保留图像的纹理细节。检测效果优于常用整数阶微分算子及现有的一些0~1阶分数阶微分算子。

Common integer differential operators for edge detection cannot keep the texture information of image well. Aiming this problem,on the basis of four-direction Roberts operator,Prewitt operator and Sobel operator,we use 0 ~ 1-order fractional differential to replace one-order differential and construct three novel 0 ~ 1-order fractional differential operators for image edge detection. Experimental results demonstrate that the three constructed fractional differential operators can effectively extract the edge characteristics,and can keep texture information of the image to greater extent as well. Its detection effects outperform that of the common integer differential operators and some existing 0 ~ 1-order fractional differential operators.