Recently, ℓ1-based image deconvolution has demonstrated superior restoration performance to other regularizers, and thus, receives considerable attention. However, the restoration quality is generally sensitive to the selection of regularization parameter. The key contribution of this paper is to develop a novel data-driven scheme to optimize regularization parameter, such that the resultant restored image achieves minimum mean squared error (MSE). First, we develop Stein's unbiased risk estimate (SURE)--an unbiased estimate of MSE--for image degradation model. Then, we propose a recursive evaluation of SURE for the basic iterative shrinkage/thresholding (IST), which enables us to find the optimal value of regularization parameter by global search. The numerical experiments show that the proposed SURE-based optimization leads to nearly optimal deconvolution performance in terms of peak signal-to-noise ratio (PSNR).
|