Texture Description Using Fractal Analysis

Texture classification is a very important research work in the fileds of pattern recognition and computer vision, and has an extensive application background in science and engineering technology. Texture feature extraction is important for texture classification. It's the fundamental job for latter classification. Although much research achievement has also been obtained, there still exist many research tasks to be resolved.In this paper, texture feature extraction and texture classification are studied. In the first chapter, the background of this paper is described. In the second chapter, the fractal theory is described. In chapter three, the general methods of texture extraction are provided. In chapter four, an SVM classifier and some general classifier are described.Before classifying texture feature, we must first extract texture feature. A texture descriptor is proposed, which combines local highly discriminative features with global statistics of fractal geometry to achieve high descriptive power, but also invariance to geometric and illumination transformations. Local fractal features are estimated densely on the filter responses. A texton dictionary is learned from the local fractal feature vectors. Then we partition the images pixel by mapping them to the texton dictionary and compute each pixel sets' fractal dimension to obtained the final texture feature MFS vector.A Support Vector Machine (SVM) classifier is used in the experiments to classify the feature. SVM which is based on statistical learning theory and optimization theory is a new classifier. SVM overcome traditional classifier's drawbacks and make a good performance. The experiments results on UIUC database show that our method achieves state-of-the-art performance as compared to fractal based methods and make superior performance than the MFS method.