A Novel Fractional-order Signal Processing Based Edge Detection Method

Edge is one of the most essential features of image. Edge detection is the elementary and funda-mental stage of image processing, and image processing applications have been integrated with various areas of human activity. How to balance the edge detection accuracy and noise immunity has always been important and difficult for all researchers. This thesis is devoted to develop a novel scheme to solve this conflict, and achieves remarkable results.On the basis of analysis and synthesis of the existing research results, the thesis proposes an novel edge detection operator, YE&YANG operator, which has two versions, i.e., first-order and second-order version. The main contribution of the thesis are listed as follows:Firstly, the thesis adopts fractional differentiation and integration to obtain so-called composite first/second-order derivatives. Then the amplitude and phase characteristics of the composite derivatives are analyzed and the advantages of the composite derivatives are illustrated.The operator is firstly realized by the frequency domain approximation method, and then its prac-tical implementation in edge detection is discussed and four schemes are proposed. One best scheme is selected through experiments.Then, the fractional order signal processing based first-order edge detection method and fractional orders signal processing based second-order edge detection method are proposed. The implementation processes of the two methods together with theoretical study, algorithm analysis, and parameter design are elaborated in details. The detection accuracy and noise immunity of the two methods are illustrated. The advantages are analyzed and highlighted.Finally, the detection results of the two methods are compared with those by the classical algo-rithms, such as CANNY operator, LOG operator, etc., using both subjective and objective evaluation criterion. The effectiveness of the methods are proved. The experimental results verify that the new operator is promising although there are deficiencies that need to be further studied.