This paper presents an automatic method to evaluate the tree crowns presented in the high resolution satellite images. This method comes into picture to reduce the manual power, time and money. Extraction of urban vegetation is a complex one. So, a couple of spectral indices called NDVI (Normalized Difference Vegetation Index) and SI (Saturation Index) are used to identify the vegetation area presented in test area. We apply thresholding techniques to identify the trees followed by edge detection using sobel operator to delineate the tree crowns presented in the test area. Finally, count the number of tree crowns by using morphological open operations (remove the noise components) and connectivity methods. The results obtained are compared with manual delineation and evaluate the accuracy in finding the tree crowns by this approach. The factors effect the performance are put forward and discussed.
