A method for detecting artificial objects in natural environments

摘要

In this paper we will present a method ft)r automatic detection of man-made objects in digital images representing natural environments. This method is based on statistical distribution of texture patterns in the image. This distribution is computed using Zipf's law. The image is divided into sub-frames and Zipf's distribution is computed for each sub-frame. Then the surfaces tinder Zipf's plots of the different sub frames are compared in order to determine which sub frame will contain an object. The sensitivity of this detection method to image resolution is also examined.