Classification Of Texture Image Based On Quaternion Wavelet Transform

Along with the rapid development of the digital image processing technology and computer application, people try to make computer process digital image automatically. These applications improve the quality of life and the efficiency of work. Analysis of texture is a very important research work in the fields of pattern recognition and computer vision, and has extensive application background in science research and engineering technology. In recent years, people do plenty of research to extract the texture features of texture image, and have obtained many achievements. Most of these studies are based on library Brodatz.In this paper, we try to extract the description of texture image in using quaternion wavelet transform, because quaternion wavelet transform can compensate the deficiency of the classic wavelet transform and obtains more information than classic wavelet transform. Because this is a new method of wavelet transform, we can barely find the papers about it. Fortunately, our experiments show that even using K-nearest neighbor classifier we can still have satisfactory results.This paper not only presents the concept of quaternion wavelet transform, also realized the application in classification of texture image, and obtains very good results. At the end of paper, two possible forward studies have been suggested.