Segmentation of nonuniformly illuminated images using adaptive windowing and nonhomogeneous granulation

In this paper, the problem of segmenting images with nonuniform lighting conditions has been addressed. It has been observed that for such images homogeneous granulation based thresholding algorithms yields poor results. In order to deal with such images, the notion of adaptive windowing has been employed using the notion of window growing. The windows have been fixed based on the entropy measure. Each window is segmented based on the proposed nonhomogeneous and heterogeneous granular computing. The results obtained by the heterogeneous and nonhomogeneous granule based scheme have been compared with homogeneous granular based scheme and Otsu's method. The results obtained by the proposed method are superior to that of homogeneous granulation and Otsu's method [3].