Abstract: This paper proposes an color edge detection scheme based on Cellular Neural Metworks（CNN） mode.In order to take full information of the color image，uses Mahalanobis distance to measure the difference between the pixels in the RGB color space.The output image of conventional methods of edge detection is binary images.To solve this shortcoming，because CNN’s output images can be multi-value image，uses CNN to color image edge detection.Mahalanobis distance is used to modify gray CNN mode.Revising its measure of a difference between the pixels，it can be operational in the RGB color space.The experimental results indicate that the new method is more in line with the perception of the human eye.In addition，in the abundant local details or some tiny changes regions，the new method can achieve better results.

Key words: edge detection, Cell Neural Network（CNN）, Mahalanobis distance, RGB color space, visual function

摘要： 利用细胞神经网络（CNN）模型导出了一种新的彩色图像边缘检测算法。为了充分利用图像中的颜色信息，在RGB彩色空间中用Mahalanobis距离来度量象素之间的差异。为了解决常规边缘提取方法输出二值结果的缺点，采用可以多值输出的CNN来进行彩色图像边缘检测。通过Mahalanobis距离对灰度CNN度量象素差异的方式进行改进，使其可以在RGB彩色空间中进行运算。通过与Sobel、Log和Canny等几种边缘检测算子比较，可以看出新方法的结果更加符合人眼的感知。此外，在含有丰富细节和微小变化的区域，新方法可以取得更好的结果。

关键词: 边缘检测, 细胞神经网络（CNN）, Mahalanobis距离, RGB彩色空间, 视见函数