Unsupervised Color Image Segmentation Based on Local Fractal Dimension
Karin Satie Komati, Evandro Ottoni Teatini Salles, Mário Sarcinelli-Filho · 2010
This paper proposes an improved version for the JSEG color image segmentation algorithm, combining the classical JSEG algorithm with a local fractal operator that measures the fractal dimension of each pixel, thus improving the boundary detection. Furthermore, the sensitivity of color variation is enhanced when working with the original color value, instead of quantized color information. Experiments with natural color images of the Berkeley Segmentation Dataset and Benchmark (BSDS) are presented, which show improved results, qualitatively and quantitatively, in comparison with the classical JSEG, the Fractal-only and the Fractal-JSEG methods. In this paper, we propose a new approach to solve such problem, enhancing the sensitivity of color variation working with the original color value, instead of a class of this color as JSEG. We continue combining the local fractal dimension in the JSEG algorithm, enhancing the detection of boundary regions, and, as a consequence, the image segmentation results. The new method thus generated is hereinafter called I-Frac .