Texture Classification and Segmentation based on Bidimensional Empirical Mode Decomposition and Fractal Dimension

In this paper, we proposed a scheme for texture classification and segmentation. The methodology involves an extraction of texture features using bidimensional empirical mode decomposition and fractal dimension, then, is followed by a k-means based classifier which assigns each pixel to the class. In feature extraction, firstly, the intrinsic mode functions (IMFs) which directly from image data by means of bidimensional empirical mode decomposition were obtained. Secondly, we calculate Differential Box-Counting of each intrinsic mode function as texture features. After feature extraction, K-means clustering is performed to the texture image. The main contribute of our approach is to using fractal dimension of each IMF as texture feature. Preliminary result, this scheme show high recognition accuracy in the classification of Brodatz texture images, and it can be also successfully applied to image segmentation.