Applying Entropy in Edge Detection from RS Images Using a Knowledge based System and Piecewise Thresholding

Edge detection is a fundamental concept of image processing in remote sensing which aims at detecting points of the image at which the brightness changes sharply. Sharp changes in the image properties usually represent important events and changes in environmental factors. This kind of detection method can be useful in different image processing applications such as satellite image segmentation. Many classical mathematical algorithms for the edge detection, such as Gradient operators, Laplacian and Laplacian of Gaussian operators,are based on the derivative of the original image pixels. In the remote-sensing imagery, because of the high rate of changes, these edge detection operators perform weakly in correct diagnosis of the feature boundaries and in keeping their consistency. In order to solve these problems, this research presents a knowledge-based system to detect the feature edges based on the properties of remote-sensing images. In this method, the threshold is firstly determined piecewisely in different regions of the image using a knowledge-based system, and then the area boundaries are extracted using the Shannon entropy, which safeguards the continuity of the edges. The advantage of this approach is simultaneous evaluation of the image features in both partial and general cases. In addition, the system is flexible in terms of the type of user needs and the intended details and applications. Eventually, to evaluate the method, the results are compared with Sobel, Perwitts and LOG standard algorithms. It is observed that the proposed technique can efficientlydetect the edges for different digital images.