基于多尺度各向异性高斯核的彩色图像边缘检测算法

Color Image Edge Detection Algorithm Based on Multi- scale Anisotropic Gaussian Filter

文中提出一种新的对噪声鲁棒的彩色图像边缘检测算法。该算法利用多尺度乘积作为提取彩色图像的边缘映射的测度。首先,分别利用单尺度各向异性高斯核的方向导数计算彩色图像的边缘强度映射后再求取尺度乘积的均方值,然后通过非极大值抑制找出候选边缘像素点,最后在高低门限值的约束下去除伪边缘像素点,通过滞后判定实现边缘连接并将边缘图作为掩码与彩色图像结合形成彩色图像边缘图。与Canny算法相比,该算法通过多尺度高斯核的结合得到较好的边缘分辨率,同时引入依赖于噪声的高低门限来控制虚假边缘的发生可能性。

A newcolor image edge detection algorithm of noise robustness is proposed. It uses multi- scale multiplication as measure to extract the color image map. First,single- scale anisotropic Gaussian kernel directional derivative filters are used to calculate image intensity map respectively and calculate the mean square value of multi- scale multiplication. Then through the non- maximum suppression,the candidate edge pixel is found. Finally,by the high and lowthreshold values and the implementation of lagging,the false edge pixels are erased to realize edges connected,combining edge figure as a mask with color image to form the color image edge. Compared with the Canny algorithm,the algorithm obtains,by the combination small and big scale of the Gaussian kernel,a better edge resolution. At the same time,high and lowthreshold dependable of noise is introduced to reduce the possibility of a false edge in noise background.