Abstract:

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.

Key words: machine vision, image processing, edge detection, grey theory, GM （ 1,1, C ） model