基于局部约束的人脸图像超分辨率重构算法

摘要

基于超完备稀疏表示理论，并根据人脸图像的特征，提出一种基于局部约束的人脸图像超分辨率重构算法。该算法首先通过样本训练出一对高、低分辨率相关联的冗余字典；再根据局部范围内人脸图像的相关性，重构出高分辨率图像；最后对图像进行全局优化。为验证算法的有效性，本文利用ORL标准图像库进行了对比实验，实验结果表明，该算法能够有效提高峰值信噪比，同时能够更好地恢复人脸图像的高频信息，有一定的实用价值。%Based on the overcomplete sparse representation theory,a face image super-resolution reconstruction methodbased on local constraint is proposed.First,a pair of associated redundant dictionary,including high-resolution (HR)and low-resolution (LR),are calculated through sample training.Based on the correlation in local scope of face im-age,the high-resolution image is reconstructed.Then,the image is optimized globally.To verify the validity of the algo-rithm,the contrast experiment is made by using ORL standard image,the results show that the algorithm can work wellin the actual conditions,and can recover the high-resolution information perfectly.