Study on an improved differential box-counting approach for gray-level variation of images

Abstract

The fractal dimension, an important parameter as a measure of roughness of image, has been widely utilized to image classification, recognition and segmentation etc. Differential box-counting approach is widely applied to estimate fractal dimension in the calculation approaches of fractal dimension. However, this approach can not accurately calculate fractal dimension of image which have smaller gray level. In response to the issue, this paper proposes an improved differential box-counting method on the height h' of box. In order to verify the superiority of the improved algorithm, DBC, RDBC, SDBC and improved DBC are separately utilized to estimate fractal dimensions of random images with different gray levels and texture image with different sizes, and then to compare. Experimental results demonstrate that: improved differential box-counting approach is more stable for random images with different gray levels.