Image deblurring via enhanced local maximum intensity prior

Abstract The enhanced local maximum intensity prior is proposed in this paper. We find that the intermediate latent image restoration can be enhanced by joining the local maximum intensity with the local maximum gradient, and the quality of the recovered kernel is hereby improved. Our method is based on the fact that the maximum value of local patch pixels and gradients will decrease with the blurring process, which has been proved mathematically and experimentally. This joint prior has been named enhanced local maximum intensity prior. Based on this prior, a new energy function is established and the image deblurring is successfully carried out by alternating optimization strategy. Extensive experiments have been conducted, which shows that our method is more effective compared with the state-of-the-art methods.