Background removal by modified fuzzy C-means clustering algorithm

This paper proposes a modified Fuzzy C-means clustering algorithm which removes the different types background present in the image automatically. It removes the limitation associated with Fuzzy C-means clustering algorithm such as selection of initial number of clusters by selecting from the co-occurrence matrix. As the co-occurrence matrix represents the repetition of adjacent gray levels, the proposed algorithm selects centroids as the gray levels corresponding to the diagonal elements of the co-occurrence matrix. Some morphological operations were done to the resultant image. The proposed algorithm uniformly removes the background pixels. The performance of the proposed algorithm is measured quantitatively.