摘要： 针对一般学习算法效率低下的问题，提出一种马尔可夫网络模型下的非线性学习算法。对输入的低分辨率图像以及训练用高分辨率图像和对应的低分辨率图像进行分块，并使图像基本对齐，构造训练图像集，利用训练集人脸图像的差异，采用块坐标限位操作技术，给出一种非线性样本搜索算法，降低搜索空间复杂度，提高了匹配效率和相关性。利用搜索到的高分辨率图像分块样本，直接输出超分辨率图像。分析和实验证实，与传统学习算法相比，该方法具有输出质量好、效率高的特点。

关键词: 人脸图像, 超分辨率, 马尔可夫网络, 非线性搜索

Abstract: This paper researches a face image Super Resolution(SR) algorithm based on learned image examples. It uses patch-based Markov network to express the mechanism of SR processing. After dividing the high-resolution images and the corresponding low-resolution ones into patches, it sets up the training dataset. Considering the requirements of Markov network computing and the difference among the images in training dataset, it proposes a patch position constraint operation for searching the matched patch and a nonlinear searching algorithm. These techniques can decrease the complexity of the searching operation and increase the effect of matching. After collecting the matched high-resolution patches, the proposed method directly uses them to integrate an output image. Experimental results demonstrate that the algorithm has a better performance and higher efficiency.

Key words: face image, Super Resolution(SR), Markov network, nonlinear search

中图分类号: 

• TP751