Color Image Segmentation of Multiple Occurrences of ROI in Satellite Images Using CCL

This study applies edge detection methods to each band of satellite color images for segmenting the region of interest. Edge detectors affect further processing and perform the basis for successful recognition of objects and classification problems. For this reason, a performance evaluation based on PSNR value, total elapsed time and the number of connected component involved in identifying the Region of Interest is explored in satellite images. To reduce the number of connected components the image is convolved with masks(3X3 laplasian, 5 X 5 mask,7 x 7 mask). The method investigated is Sobel, Prewitt, Log, Zero Cross and Canny detectors. Among the methods investigated Sobel and Prewitt produces good representative edge detection for satellite images. Then the detected area is superimposed to show the objects clearly by smoothing that area with Gaussian filter and divide it d with input image.