基于自适应规整化方法的图像超分辨率重建

Image super-resolution restoration based on adaptive regularization MAP

以Gauss-Gibbs随机场模型为图像的先验概率模型,运用自适应规整化的最大后验概率(MAP)方法进行图像超分辨率重建。通过对先验概率分布参数的估计,对图像超分辨率重建求解进行自适应规整化,从而提高重建图像的质量。实验结果表明,该算法能较好地再现图像的各种边缘信息,重建的高分辨率图像在峰值信噪比和视觉效果方面都得到明显提高。

A MAP method based on adaptive regularization was proposed for image super-resolution restoration, the prior distribution of the image was assumed in Gaussian-Gibbs field. The solution was adaptively regularized through parameters estimating of prior distribution. Experiment results show that the method can properly retrieve all kinds of edges. In the meanwhile, the PSNR and visual quality of restored high resolution image are improved significantly.