The Study Of Image Retrieval Technique Based On Wavelet Transform

The content-based image retrieval technique plays a crucial role in multimedia application. The compressed domain retrieval methods reduce the need of the computer;improve the real time of computer and agility. These impelled the popularization and developing of multimedia. So the content-based image retrieval techniques were used more and more popular and widely, retrieval techniques based on compressed domain becomes a new important trend of CBIR.In the dissertation., the investigation work has been done around techniques of image retrieval based on compressed domain. The main content are as follows:(1) Analyzed the compressed domain retrieval methods at present, summarized the key technology in the image retrieval technique.(2) Analyzed the statistical attribution of wavelet transformation coefficients. The experiment shows that an image retrieval method based on wavelet domain was proposed by using Gaussian mixture model of wavelet coefficients distribution. The effect of using Gaussian mixture model is better than generalized Gaussian model.(3) Improve the wavelet-based image retrieval which based on generalized Gaussian. We described approximately the coefficients distributing by using Gaussian mixture model to fit wavelet coefficients histogram. The experiments show that the retrieval effect of using Gaussian mixture model is better than generalized Gaussian model.