Sub-fractal Dimension Analysis And Its Application In Image Segmentation Of Seismic Sections

Texture feature analysis has been one of the most important but difficult properties for texture image segmentation. It is important because it provides the essential structure information in different regions of a texture image. Fractal and wavelet process inherently multiple-scale property and scale invariant that can be competent for roughness features in textural representation of texture image.We focus on Fractal Dimension (FD) feature extraction method of fractal-based theory for texture image segmentation in this dissertation. We first study general methods of estimating fractal dimension in image analysis. And then, we propose an improved method of estimating fractal dimension. Furthermore, we present a new approach of estimating fractal dimension, which related measure employs the singularities test theory of the locations maxima of the wavelet transform modulus in signal processing field.For the purpose of reducing estimating error in segmentation, we use the 8-neighborhood edge-preserving noise smoothing quadrant filter (EPNSQ), instead of the traditional 4-neighborhood technique, to smooth each of the fractal dimension features before practical segmentation. An iterative K-means scheme is used for segmentation. Our segmentation test runs in the particular images of real seismic section. The effectiveness of our estimating FD approachs and the proposed segmentation algorithm are confirmed through computer simulations.