An improved Canny Arithmetic based on image pyramid theories

Canny Arithmetic Operator has been proved to be a traditional operator possessing a better detective effect in the common usage of edge detection algorithm currently. However, Canny Operator also has certain deficiencies, such as choices of dimensions and Threshold Limit Values. When the dimension is bigger, effect of noise to the edge is smaller, but the locating accuracy will be lost. Whereas, when the dimension is smaller, the locating accuracy will be high, but the ratio of Signal-to-Noise of output image will be lower. In the meantime, using a same Threshold Limit Values (TLV) in the whole image would also influence the edge detection in dissimilar districts. This paper presents an improved algorithm to the edge detection, which uses image pyramid theories to improve the Canny Operator. It is also demonstrated in this study that this method can rectify the shortage perfectly, and obtain a more ideal edge image.