基于MAP技术的图像类推超分辨重建方法

MAP-Based Image-Analogies Super-Resolution Reconstruction

在最大后验概率(Maximum A Posterior,MAP)的基础上,结合图像类推(Image Analogies,IA)思想,提出一种序列图像超分辨率重建方法——MAPIA(Maximum A Posterior Image Analogies)。该算法先利用传统MAP方法将序列图像进行超分辨率重建,然后在序列图像中选取一帧图像与重建后的图像构造训练集合的图像对,学习它们之间的关系,利用图像类推技术进行超分辨率重建。实验证明文中方法不仅能有效提高图像的清晰度,而且较其它的方法,能得到边缘更加清晰、细节更加突出的重建图像。

Presents a new method（Maximum A Posterior Image Analogies,MAPIA） based on MAP for image super resolution in use of image analogies. Firstly, use the Bayesian MAP algorithm for supper resolution image reconstruction; secondly, use a low resolution image and it＇s super- resolution version to construct the training set. Learn the relationship between source image pair, and then,tit this problem to image analogies for super resolution. The experiments show the approach can amplify images,and the edges in the MAPIA reconstruction result are much elearner and the details are more prominent over traditional methods.