FRACTAL-BASED TEXTURE ANALYSIS

Texture analysis hai been widely applied in image segmentation, image classification, pattern recognition, etc. One approach is to extract the textural information from the region of interest by calculating first-order or second-order statistics. Many studies have found that spatial grey level co-occurrence matrices statistics are good textural parameters. However, they are not particular effective in the characterization of textures which are of low contrast and random in nature.
Recently, another approach to texture analysis begins to emerge, which is based on fractal geometry. The mathematical model of fractal geometry has been used to analyze many natural and physical phenomena. In this investigation, it was applied in the modelling of natural texture exhibited in digital image. One important parameter that can be derived and then used for texture characterization is the fractal dimension, which can be determined using two separate methods. In the first method, the fractal dimension is estimated based on the average intensity difference of the pixels within the region of interest. In the second method, the fractal dimension is determined from the Fourier power spectrum of the image data. In order to assess the effectiveness of the fractal dimension, the regions of interest were also analyzed using spatial grey level matrices statistics.
Fractal dimension seems to be good in describing many natural textures as well as many human tissues as displayed in medical images such as the ultrasonic image. The results also show that the fractal dimension is a better parameter than many secondorder statistics in texture characterization. The possible reasons are that many natural textures are very similar to fractal surfaces, and also the way the fractal model describing surface roughness correlates closely to human perception of texture. © 1992 IEEE.