Features Extraction And Classification Of Rice Paper Based On Wavelet

Chinese traditional painting is very special for Chinese nation. Because of its unique history and rare works, many people or museums join the collection of traditional Chinese paintings. However, there are many fakes on the market and it is a must to check on identification of Chinese traditional painting. As the main carrier of traditional painting, rice paper is rich in fiber texture. Up to present, there are still few progress on the texture discrimination of rice paper. So we will do a research on how the wavelet transform effects the classification rate of texture features extracted form images of rice paper.I will carry out thesis in three ways. The first part is mainly about the analysis of texture of rice paper. I introduce the general situation of research and discuss the four methods of texture application.The second part mainly expounds the basic theory of wavelet function and the classical wavelet transform. And then I did the classification experiments on rice paper texture image using wavelet transform. The experimental results show that the paper texture classification using the DB2 wavelet transform can achieve the highest accuracy of 87.6%.The third part mainly introduces the basic theory of the gray level co-occurrence matrix. And then I studied the influence of structure factors on rice paper texture characteristic parameters and determine the structure factors for rice paper texture values by the experiment.Finally, I combined of wavelet transform and gray level co-occurrence matrix this two kinds of methods, to carry out a experiment to classify texture of rice paper image. The experimental results show that using the bior1.1 wavelet and gray level co-occurrence matrix to extract features of rice paper, the classification rate is the highest, is 94.7%, and the time is short.I used wavelet transform to extract features of rice paper and proposed a new method, combining wavelet transform and gray level co-occurrence matrix to classify texture features of rice paper. This method can successfully discriminate 6 kinds of texture images of rice paper. The experiments show that the combination of this twomethods can extract features by not only using the statistical method but also using signal processing method. This new method can greatly improve the accuracy of classification of rice paper.