PERFORMANCE ANALYSIS ON EDGE DETECTION ALGORITHMS FOR COASTLINE IMAGE DETECTION

摘要

Edge detection is one of the most important areas in the image analysis. Since edge detectors are not at ease applying to different situations, it would be ideal to find out edge detectors that work best in analysing satellite images. In this paper we evaluate the performance of commonly used edge detection algorithms for extracting coastlines over satellite images. The edge detection algorithms applied in this study are: Prewitt, Sobel, Canny and Morphological method. A labeling method is applied to the images before the edge detectors are used for coastline detection. Our comprehensive evaluation on the performance of these algorithms in detecting coastlines over the satellite images suggests that the Sobel edge detector offers the best performance. In the paper to measure the quality of the coastlines detected over the processed images after pruning, the Peak Signal to Noise Ration (PSNR) and Root Mean Square Error (RMSE) parameters are applied upon the images generated by each edge detection algorithm.