一种用于轮廓线探测的CNN改进算法

Modified CNN Algorithm for Contour Detection

图像中目标物体的轮廓探测是目标识别和计算机视觉系统的第一步也是关键一步。提出了一种基于细胞神经网络(Cellular Neural Network,CNN)的轮廓线探测改进算法。该算法中CNN模板参数(模板系数)是根据局部窗口内各像素与中心像素间的灰度和空间关系计算的,即模板参数的计算不仅考虑了局部窗口内各像素与中心像素的灰度值差异,而且顾及了窗口内各像素与中心像素间的空间距离。实验结果表明,相对于其它两种轮廓探测算法,提出算法的探测效果较好。

Contour detection of object from image is the first and crucial step in computer vision and object recognition system. A modified contour detection algorithm was proposed based on cellular neural network （CNN）. In proposed algorithm, the template parameters （template coefficients） of CNN were calculated according to the gray-scale and spatial relationship between the central pixel and the other neighboring pixels in the current local window, i.e., computation of template parameters considers both gray value difference and spatial distance between the central pixel and the other neighboring pixels in current local window. Experimental results show that the proposed algorithm has better detection performance than other two contour detection algorithms.