基于灰模型白化响应的边缘检测算法

Edge Detection Algorithm Based on Grey-Model Whitening Response

为获得更好的图像边缘,提高机器视觉的检测精度,提出一种新的基于GM(1,1,C)模型白化响应的图像边缘检测算法.该算法用原图相邻的若干像素点值构建GM(1,1,C)模型,计算出相应像素点的白化值,得到原图像素点的像素值与相应白化值之间的误差.依据边缘像素点的像素值与非边缘像素点的像素值相差大而不满足GM(1,1,C)建模条件,从而导致边缘像素点白化值出现较大误差的特点,实现图像的边缘检测.实验证明:该算法在无噪声和有噪声的情况下均有效;建模像素数量越少,抗噪能力越强,但对边缘的检测能力会下降;建模像素数量越多,边缘的检测能力越强,但抗噪能力下降.

In order to accurately detect the image edge and to improve the detection precision of machine vision, a novel edge detection algorithm based on the whitening response of GM （ 1,1, C ） model is proposed. In this algorithm, the neighboring pixels of the original image are used to establish a GM （ 1,1, C） model for the calculation of the corresponding whitening values. Thus, the errors between the whitening values and the original pixel values are obtained. As the edge pixel values are different from the non-edge ones, the modeling condition of GM （ 1,1, C） model can not be successfully satisfied and a large error in GM （1,1, C ） whitening value may occur, which makes it easier to effectively detect the edge. Experimental results indicate that the proposed algorithm is effective in both the noisy and the non-noisy conditions ; and that, with the decrease of the pixel number for modeling, the anti-noise ability of the algorithm improves while the detection ability for image edge decreases. However, opposite results are obtained with an increasing pixel number.