Multiresolution Edge Detection using Particle Swarm Optimization

Abstract

In this paper, a heuristic approach based on biologically inspired Particle Swarm Optimization (PSO) is proposed to be used at multiresolution level to improve the quality of detected edges. In the preprocessing stage, one level of Discrete Wavelet Transform (DWT) or Dual-tree Discrete Wavelet Transform (DT-DWT) is applied to the input image to create new subbands images. PSO is then applied to each one of these subband images. The output image containing the detected edges is obtained by reconstructing it from the processed subband images using the inverse transform. An objective function is proposed for the PSO to evaluate edges during the heuristic search within the image space. Also, automatic thresholding is introduced which is used to automatically threshold the output of the PSO into binary image. Performance the proposed approach is evaluated and compared with other well-known edge detectors such as Sobel and Canny using Kodak image database. The results from objective evaluation using Peak Signal-to-Noise-Ratio (PSNR) and Root Mean Square Error (RMSE) showed that the proposed approach has a better and/or comparable performance compared to other edge detectors.