Thresholding for edge detection using fuzzy reasoning technique

Threshold values for extracting effective sketch features are different according to images. Moreover, a global threshold cannot extract sketch features effectively because of the fact that there are various types of objects and regions having many different gray levels within a single image, and the human visual characteristics are more sensitive to the sketch features in dark regions than in bright regions. Therefore, in order to extract sketch features in a manner based on the human visual characteristics, this paper suggests a method for deciding whether a pixel is a feature or not, deduced through fuzzy reasoning utilizing a contrast and a local brightness as its input measures. To do this, a contrast and a local brightness are proposed and these membership functions are generated subject to the histograms which are constructed from applying these measures to the input image. In order to decide the degree of threshold values according to image context, an output membership function is generated subject to the standard deviation of input image. Several experimental results are presented to support the validity of proposed method