Edge Detection using Ant Colony Optimization under Novel Intensity Mapping Function and Weighted Adaptive Threshold

Abstract

Edge detection is a crucial phenomenon in image segmentation. In general, kernel based methods like Sobel, Canny, Roberts etc. are used which are based on first and second derivatives pixels intensity. However, these methods fail to find all the true edges. Moreover, number of falsely detected edges is much more than true edges. This happens due to a fixed threshold used in these methods. To reduce falsely detected edges, a method which can dynamically adjust its threshold is desirable. Artificial and swarm intelligence based methods are capable to handle minute details. In this work, Ant Colony Optimization (ACO) based method is detailed for edge detection. In this method, a novel function is used to capture intensity variation in a particular image. In learning based method adjustment of threshold is also necessary to obtain good results. In this work, we have considered weighted average for threshold update in contrast to earlier method where simple average is taken. Finally, the performance evaluation and comparison is made in terms of Peak-Signal-to Noise Ratio (PSNR), accuracy and F-Score and usefulness of proposed method is shown.