Based On The Total Variation Of Noisy Blind Image Restoration Algorithm

During the process of image acquisition, transmission and display, inevitably the image degradation will appear and express as blurring and noise pollution. Image restoration means dealing with the observed degraded image and restoring the original image with the greatest fidelity. Traditional image restoration methods are based on the assumptions that the PSF (Point Spread Function) of the system and noise distribution have been acquired, and then use the anti-degradation process to recover the original image. However, in real practice, prior knowledge of images and imaging systems is completely or partially unknown;under this circumstance, the process of restoring the original image only according to the observed degraded image is called blind image restoration. It is a much practical and challenging subject.In this paper, the two main aspects of noising image blind restoration (de-blurring and de-noising of image) have been researched in depth; it also elaborate the image degradation model, various image blind restoration and de-noising methods. Based on the superiority of keeping edge information and de-noising about the total variation (TV) image restoration algorithm, it critically focuses on this algorithm by deeply analyzing the model, feature, numerical realization and a variety of improved forms. Moreover, based on the analysis above, due to the drawback of "staircase" effect from TV blind restoration algorithm, it proposes an improved algorithm which inserts the anisotropic LPA-ICI de-noising algorithm to the iterative procedure of blind restoration algorithm, ameliorating the scales selecting for anisotropic LPA-ICI de-noising algorithm, thereby showing an adaptively updated method for the improved algorithm. Meanwhile, combining with the iterative blind deconvolution idea, this paper interactively using TV blind restoration algorithm in the field of frequency domain and spatial domain, and limiting the image and PSF via priori knowledge, simplifying the choice of the regularization parameter, therefore reducing the complexity of the algorithm. Last but not the least, in order to gain a better restoration results, a multi-channel image blind restoration algorithm has been introduced by extending the single-channel TV blind restoration algorithm to multi-channel; and motion estimation was applied as a pre-procession to the multi-channel image restoration process to remove the distortion of the image motion warping, thus actualizing the multi-channel image blind restoration algorithm which is based on motion estimation. To verify the validity of this result, several groups of blind restoration experiments for the degraded images had been tested. From the experiment of single-channel, it obviously remove the drawback of "staircase" effect then improved TV blind restoration algorithm, which performs better on aspect of de-noising and keeping edge information. In addition, the experimental results of multi-channel also shows that the multi-channel image blind restoration algorithm could gain effective result rather than fusing restored images achieved by restoring the single-channel image separately. Regardless of whether the collection condition is ideal or the motion slightly warps, the algorithm could restore the degraded image in high quality and has good stability and astringency.