Abstract: Because of the good performance in maintaining the target profile and partitioning the object from low-contrast pictures,threshold segmentation method based on edge information is widely used and especially suitable for industrial production images.But traditional methods are sensitive to noise,and the threshold is hard to select.To solve these problems,in this paper,an adaptive image threshold segmentation algorithm based on edge information was pre-sented.The proposed algorithm uses characteristic response of SUSAN to describe the edge information of pixels to suppress the effect of noise and weak boundaries.Time and space complexity is greatly reduced when using the min-max cut threshold segmentation algorithm based on graph spectral theory rather than other segmentation algorithms,and the received threshold is global optimum.Experimental results show that the algorithm can segment the target accurately and retain rich details at the same time,for low-contrast and noise images the algorithm also has good performance,and threshold is better than traditional algorithm.

Key words: SUSAN operator,Edge information,Graph theory,Min-max cut rule,Image thresholding segmentation