An Algorithm of Extraction Visual Texture Features Based on Gabor Transformation

A kind of algorithm to extract global texture coarseness and direction based on multi-scale and multi-directional metric model of Gabor filter is proposed. In this method, at first, Gabor wavelet model based on multi-scale and multi-directional transform is constructed and image filtering is executed on three-scale and four-directional metric model of Gabor filter. 12 fractal dimensions are extracted based on filtered 12 images, and the average of fractal dimension is calculated as presented the model of measuring global texture coarseness. Then a histogram of local edge probability is constructed to describe qualitatively the distribution of global texture and give a quantitative value of measurement. At last, the distance matching is performed to obtain a similarity measure of combination textures. The proposed method is employed in an image retrieval system using combination visual texture features representation. It shows that the method is efficient.