A new box-counting method for image fractal dimension estimation
Song Xue, Xinsheng Jiang, Jimiao Duan · 2017
A fractal dimension is an effective feature for texture analysis, segmentation and classification in many fields. The most frequently used method to estimate the fractal dimension of an image is the differential box-counting method. This method is simple but not accurate enough. Many researches have been done to improve its counting accuracy. So far, the counting accuracy along z direction has been improved more effectively than along x and y directions. In this paper, some typical improved methods are summarized and analyzed. And then a new method based on the idea of dilation is proposed, which is simple and understandable, and has the same desired counting accuracy in x, y, and z directions. Simulation results demonstrate our method can outperform other existing methods in estimation accuracy and robustness.