Image restoration algorithm for multiple degradation caused by blur and mixed noise

点击翻译按钮获取中文摘要

Total Variation (TV) filter is a nonlinear filter that is suitable for Gaussian noise removal, and has an advantage of preserving edges of the objects. We extended the framework of TV filter, so that it can be applied not only for Gaussian noise but also impulse noise. In this research, we applied this framework for the restoration of images degraded by the mixed noise that is composed of both Gaussian noise and impulse noise. In the proposed framework, it is required to detect spatial positions of impulse for removing the mixed noise successfully. Therefore, we introduced one of the most efficient impulse noise detectors we proposed, which is based on Gaussian curvature of the image surface. Through computer simulations, we demonstrated that the image quality of the restored images is effectively improved, compared with the conventional filters such as original TV, DW-MTM and the hybrid of switching median and bilateral filter.