For Image Denoising Perona - Malik Improvement And Study Of Numerical Algorithm Of The Model

Image denoising is a classical problem in image processing.With the development of technology of image processing,images are transmitted by the channel or preserved by the medium. Images are confined to the external physical conditions,the noise will affect the image in vision.In many application areas,we need clear and high quality images.Therefore, image denoising is not only an important image processing problem,but also an important pretreatment process for other image processing,which bring great impact on subsequent processing.The principle of image denoising is to use the difference of noise signal and the image signal in frequency domain on the distribution.The image signal is mainly distributed in the low frequency region,and the noise signal is in high frequency region due to its weak correlation with the around pixels. Traditional methods of denoising are typically based on a low pass filter, which filter out high frequency component to achieve denoising.But these mutations will affect the image visual effect.How to filter the image noise and maintain better the image details of edge and texture become the core issue of image denoising.The methods of image processing based on partial differential equation (PDE) have the anisotropic properties, strong adaptability, which can smooth noise and preserve detail information of edge and texture better. So in the past twenty years the methods have obtained huge development.The main content of this dissertation is to improve Perona-Malik model and give numerical algorithm.In the introduction,there is a brief introduction on the basic knowledge of image,the history and significance and the evaluation criteria of image denoising.In the second chapter, several common models of image denoising are introduced,such as thermal diffusion model,Perona-Malik model,regularization of the Perona-Malik model, forward-backward diffusion model,diffusion tensor Weickert model,TV model and adaptive regularization variation model based on diffusion tensor. In the third chapter, based on Perona-Malik model and Catte regularization model and considering second derivative we establish a new model of image denoising, prove the posedness of the model, and give the numerical algorithm and numerical simulation. The experimental results show that the model not only can effectively remove the noise, but also can enhance and keep the edge. In the fourth chapter, we join the intensity factor to the model, and do the numerical simulation. From the simulation results we can see that the sharpness and contrast of the processed image have greatly increased, and look more real and natural.