Study On Texture-Based Image Retrieval Technology

As technologies of computer and multimedia are rapidly developed and widely applied, the fields of publishing industry, medical and health work, architectural and engineering design produce lots of digital images every day. How to analyze, store and retrieve these images is an imperative problem. The content-based image retrieval technology can effectively resolve the problem and it also becomes a hotspot in study. The texture primary feature of image and an important method in content-based image retrieval system, so the paper mainly focuses on the texture-base image retrieval technology.Firstly, the background and importance of the content-based image retrieval was illuminated. Then, the nowadays study situation and hotspot was introduced to us. At the same time, the native application situation and overseas application situation of content-based image retrieval system were explained briefly. Lastly, some representative application system of content-based image retrieval was list in this paper.Secondly, paper researched the theory of content-based image retrieval system and discussed the frame and key technology of content-based image retrieval system. Study present: the key factor that influences the effect of image retrieval is how to describe the content of image and to measure the difference between tow images.Thirdly, paper discussed how to describe the content of image. The main job is to find an effective way to present the texture feature of image. According to the character of image texture in special and frequency field, three methods, special method, frequency method and special united frequency method, were usually adopt to present texture feature of image. This method was interpret and compared deeply in the paper. Human's vision system is sensitive to not only space but also frequency, so presenting the texture feature of image in special and frequency field has defects. The spcial united frequency method can overcome these drawbacks. At the same time, it can get a good local character in spcial and frequency field.Finally, paper explored the availability of presenting the feature of image texture in special united frequency field and researched how to finish the job by wavelet transform technology. A new method for image retrieval is proposed in this paper, which utilizes the multi-direction information of sub-bands of original image after wavelet transform. After image was decomposed by wavelet transform, there were tree sub-bands by traditional wavelet transform and six sub-bands by complex wavelet transform. The sub-bands by traditional wavelet transform included horizontal sub-band, vertical sub-band and diagonal sub-band. The sub-bands by complex wavelet transform included±15 degree sub-band,±45 degree sub-band and±75 degree sub-band. The method utilized different direction sub-band information between the sub-bands of traditional wavelet decomposing and of complex wavelet decomposing to unite sub-bands information to present texture feature. Experiments show that, comparing to complex wavelet transforms or the traditional wavelet transform, proposed method enhance the retrieval rate. In addition, paper also proposed a novel method to identify offline handwriting-based writer. After preprocessing original handwriting image, the method can identify the writer by analyzing the handwriting feature by complex transform. Our experiments show the new method, compared with the traditional wavelet–based GGD method, and our method achieves a better performance.