Texture measures for building recognition in aerial photographs

In this paper we examine the suitability of the fractal error operator for the detection of man-made features in monocular, greyscale aerial photography and propose an alternative operator, called the homogeneous operator. The fractal error operator uses a metric known as fractal error while the homogeneous operator is based on a principle of homogeneity. We have implemented and tested both operators on various images. In comparative studies, we show that the homogeneous operator outperforms the fractal error operator for images of typically rural dwellings. An image processed by this operator can be used as starting point in a building recognition system.