Abstract: A new segmentation method is proposed, in which morphology of multi-direction structuring elements is combined with mutual information technique. Initial threshold can be chosen by using maximal distances between classes and morphology of multi-direction structuring elements, and in the iteration process, an optimal threshold will be determined by maximizing mutual information between segmented and original image. It is applied to segmenting pavement images. The experimental results indicate that this revised algorithm has not only a good ability to segment the image, but also a strong ability to suppress the noise in the image. Moreover, it can provide good image processing speed.

Key words: computer application, image segmentation, multi-direction structuring elements, mutual information