A New hybrid Method for Noise Robust Estimation of Image Fractal Dimension

This paper presents a modified model to calculate the fractal dimension of digital images. The estimation of fractal dimensions is crucial to fractal analysis and is popularly carried out through methods based on box counting. The problem with these approaches is that, most of them do not remove the potential effects of noise on fractal dimensions properly. Accordingly, this study examines the effects of three different type of noises on fractal dimensions using different images taken from Background image database. The examination shows that the fractal dimensions change significantly, after noise adding, so we put forward a noise-robust and efficient fractal dimension calculation method which is a combination of two methods, the gray-level co-matrix algorithm and improved box counting method. The results of experiments on the Background image dataset confirm the robustness and efficiency of the proposed method.