Iterative Shrinkage Thresholding (IST) algorithm has been becoming popular in image restoration as well as signal processing communities. Analysis and experiments expose the low efficiency and poor robustness of IST. In the paper, we devised a Mixed Iterative Shrinkage Thresholding (MixIST) algorithm which bears more robustness, higher efficiency and better efficacy. By the power of two step iteration, advanced pre-filtering, better shrinkage methods, unbiased thresholding and the increased detail information of the original image by redundant wavelets, MixIST achieve a further improvement of the quality of restored image. In order to show the actual performance of MixIST, results from experiments on image restoration as well as comparisons with other algorithms are demonstrated.