Image Edge Detection Based on Interval Type-2 Fuzzy Similarity

This paper presents a new method for edge detection of color images that is based on interval type-2 fuzzy similarity. Firstly, the noise of the image is estimated, and the image with noise higher than the threshold value is filtered out. Then this paper combines the type-2 fuzzy set theory to calculate the adjacent pixel similarity of each channel pixels in the image. After type reduction, it is divided into the edge and non-edge pixels by threshold. Compare to Sober operator, Canny operator, LOG operator and type-1 fuzzy similarity, the proposed method has a great balance noise resistance and edge detection accuracy.