Research On Image Denoising Based On Wavelet Transform

There is a lot of noise in images. In order to analyse images, the noise needs to be reduced in images pre-processing. Wavelet analysis is a valid analyzing means. Recently, with the development and improvement of wavelet theory, wavelet analysis has been applied to many fields. At the same time, wavelet theory has been applied to image denoising successfully,too, and many new image denoising algorithms is formed.The dissertation introduces the basic theory of wavelet at first, then discusses the image de-nosing by wavelet. The research contents of this dissertation consists of three aspects: The first aspect is the researching on wavelet shrinkage threshold and proposing a new shrinkage threshold. The second aspect is the researching on relationship between wavelet coefficients and at the same time a new denoising algorithms based on Inter-Level Dependencies is introduced. Third aspect analyzes Modulus Maxima theory.Donoho has proposed the wavelet shrinkage threshold ,which is not optimal threshold ,but the maximum of the optimal shrinkage threshold. And it is irrelevant with the image itself, is only relevant with the image noise. We proposed a new wavelet shrinkage threshold which is relevant with the singularity of the image. The denoising results shows it is more optimal.There exits great relationship between wavelet coefficients. There are inter-level and intra-level dependencies, we consider the inter-level dependencies of the wavelet coefficients , a new inter-level dependencies model based on wavelet shrinkage denoising algorithms is presented. Shrinkage threshold of each subband of each level is selected according to the energy of each subband in each level and relationship of coefficients between levels, the experiment results show the new algorithms is more effective than classical one.When wavelet transformimg, coefficients of signal and noise is different on diffusing property. We can recognize coefficients of signal and reconstruct the images by the coefficients, the experiment shows the new algorithms also gains effective results.