基于超拉普拉斯先验的图像非盲解卷积

自然图像梯度的重尾分布在去噪、去模糊以及超分辨率等问题中已经被证实是一个非常有效的先验。自然图像梯度的长尾分布可以极好地用超拉普拉斯(Hyper-Laplacian)分布模型化。基于空间变化点扩散函数提出了一种超拉普拉斯先验的图像快速非盲解卷积方法,对图像进行复原。该算法的优势就是引入空间变化点扩散函数,使用分步交替迭代最小化方法,通过查表法快速对图像去卷积求解。该方法的解卷积速度得到较大幅度的提升,而且对模糊图像有较好的复原效果,提高了图像的质量。

Non-Blind Deconvolution of Images Based on Hyper-Laplacian Prior

The heavy tailed distribution of the natural image gradient has been proved to be a very effective priori in denoising,deblurring,super-resolution and so on.It can be well modeled by Hyper-Laplacian distribution.In this paper, we introduce spatially-varying point spread function (PSF),and propose fast non-blind deconvolution method based on Hyper-Laplacian prior.The advantage of Hyper-Laplacian is the introduction of spatially-varying point spread function, and it is used to solve the problem of image deconvolution by using the alternating iterative method.This method can be used to improve the quality of the reconstructed image,and it can improve the quality of image.