Study On Total Variation Method For Image Restoration

Image restoration focuses on removing or reducing the degradation which arises during the image acquisition and transportation, making the image restored get close to the ideal undegraded image, the purpose of which is to reconstruct the original image.This thesis study image restoration which is obtained under the frame of variation functional, and discuss the fine characters of the Total Variation measure for image restoration (especially for the edge), to get a better result. The essence idea of restoration based on Total Variation is that translate the problem into minimum of energy functional with TV regularization, then the Partial Differential Equations are got from the principle of variation, finally after discretization the optimal resolution with the numerical methods is obtained. The principle of TV will be illustrated in the thesis.The TV method is applied to solve two problems such as denoise and deblur, which are often concerned in the image restoration. As for the denoise, the drawbacks of traditional regularization is first discussed,and then the Total Variation's characters of suppressing noise as well as preserving the edge is illustrated, the numerical model is given in the end. The final experiments have verified the algorithm's efficiency. As for image deblur, restoring the image with known PSF is first discussed, then we expend it to a more ill-posed case, which is called blind restoration. Its target is to restore the degradation image without any information about PSF. In the blind restoration, the AM (Alternative Minimization) method is applied to decompose the process of solving PDEs into two phases such as PSF identification and image restoration. The characters of diffusion's parameters are also discussed, and an adaptive adjusting mechanism in the AM process is introduced which can preserve more detailed information. The numerical experiments have showed that the TV method has a good effect for the image restoration...