Texture Image Segmentation Based On Fractal Dimension

Based on a research of several fractal models and fractal dimension calculation methods for different fractal models, difference box dimension method is taken to calculate fractal dimension of texture Image. In calculation of the box dimension, least-squares algorithm is adopted to estimate fractal dimension for different box number and radius.As the complexity of texture image, solely using fractal dimension of original image can not lead to ideal segmentation results. To obtain more information, original image is transformed for many times to acquire direction information including vertical, horizontal and diagonal and roughness information. Man-made objects in complex background can be segmented by using fractal dimensions which consist of information above as characteristic quantities, which obtains ideal segmentation results and retains more image details. Compared with segmentation result segmented by OTSU, image segmentation based on multi-characteristic fractal dimension can retain more image details, and segmentation result is superior to threshold segmentation result.