Fractal Features for Texture Analysıs

This paper discusses the application of fractal geometry in image processing and texture analysis, specifically focusing on the computation of fractal dimension (FD). FD is a key feature that measures the level of roughness at multiple scales and is used to quantify the complexity of images. The standard method used to compute FD in greyscale images is the differential box counting (DBC) technique. However, this method has the limitations of over-counting and under-counting boxes, which can affect the accuracy of FD computation. To overcome these limitations, this paper proposes a new method that improves the accuracy of FD computation. The proposed method is tested and compared with the traditional DBC using simulated images, Brodatz texture images, and medical images from the Mini-MIAS database. The results show that the proposed method produces good results with less computational error in terms of accuracy.