Research On Image Denoising Approach Based On Statistical Model Of Wavelet Coefficients

In the process of the acquisition,, transmission and storage, digital image is often interfered by various noises, which does not only seriously affecting the visual result of the image, but also causing difficulties to the subsequent processings such as edge detection, image segmentation, feature extraction, object tracking and pattern recognition. Therefore, image denoising is a very important job in image pre-processings. With the attentions and researches of scholars in recent years, a new time-frequency analytical method--wavelet transform has been applied successfully in image denoising field. By right of multi-scale, multi-resolution analysis characteristics, the transform can provide a new and powerful method to signal processing. Based on statistical characteristics of wavelet coefficients, the paper proposes a new model-based image denoising methods, and the specific work as follows:Firstly, the image denoising technology development and research is reviewed. Starting from the principle of image denoising, this paper gives a more systematic introduction to the classification of the image denoising methods, and focuses on the wavelet-based image denoising method development. Meanwhile, the image noise model and image quality evaluation system is intimately elaborated, and the methods used in the experiments are pointed out.Secondly, taking the bivariate denoising model for the represent, the paper has a further study in wavelet statistic coefficients of denoising algorithm. It not only discusses the traditional wavelet transform theory which takes the double-density dual tree complex wavelet transform for represent, but also discusses the new multi-scale geometric analysis tool which takes contour wave transform and multiple orientation filters (PDTDFB) Transformation for present. On this basis, combining with Bayesian estimation theory, two new bivariate denoising method is presented, they are respectively the method based on double-density dual tree complex wavelet transform and PDTDFB within the framework of Bayesian maximum posterior estimation theory.There is a comprehensive conclusion to the advantages and disadvantages of new method by analyzing algorithm theory, structure and the simulation results. Comparing to the typical image denoising algorithm of home and abroad, it shows the new algorithm is efficient in suppressing image noise and keeping the image edge and texture information, so the algorithm is reasonable and effective.Finally, there is a summary of the full text, and it also takes a prospect of image denoising direction.