Fractals Based Clustering for CBIR

Abstract— Fractal based CBIR is based on the self similarity fundamentals of fractals. Mathematical and natural fractals are the shapes whose roughness and fragmentation neither tend to vanish, nor fluctuate, but remain essentially unchanged as one zooms in continually and examination is refined. Since an image can be characterized by its fractal code, a fractal code can therefore be used as a signature of the image. Image clustering supports the hypothesis that semantically similar images tend to be clustered in some feature space. The meaningful clustering is in pursuit of search for nearest neighbor in terms of similarity of the images. The objective of this work is to evaluate the use of fractal dimension as a quantitative index and effectiveness of clustering approach for image retrieval mechanism. The image retrieval mechanism has been implemented using clustering and Hausdorff dimension based fractals so as to combine the advantages of both the approaches. The results are encouraging enough to investigate use of fractals for CBIR.