有监督多类字典学习的单幅图像超分辨率重建

Single Image Super-resolution Reconstruction via Supervised Multi-dictionary Learning

针对目前基于字典学习的图像超分辨率重建算法重建效果欠佳或重建耗时较长的问题,本文提出一种基于有监督的KSVD多类字典学习算法和使用类锚定邻域回归方法来重建低分辨率图像。首先使用高斯混合模型对训练图像块进行聚类,然后使用KSVD算法,在生成子类字典的同时产生一个线性分类器;最后利用此线性分类器对输入的测试特征分类,根据相应的类字典,使用类锚定邻域回归方法来完成图像重建。实验表明,本文算法与一些经典的算法相比,在主观视觉和客观评价上都获得了更好的结果,且对人脸图像具有更好地适应性。

In order to overcome the problems that the dictionary training process is time-consuming and the reconstruction quality couldn＇t meet the applications, we propose a super resolution reconstruction algorithm which based on a supervised KSVD multi-dictionary learning and class-anchored neighborhood regression. Firstly, the Gaussian mixture model clustering algorithm is employed to cluster the low resolution training features; Then we use the supervised KSVD algorithm to generate each subclass dictionary and a discriminative-linear classifier simultaneously; Finally, each input feature block is categorized by the classifier and reconstructed by the corresponding subclass dictionary and class-anchored neighborhood regression. Experimental results show that our method obtains a better result both on subjective and objective compare with other methods, and has a better adaptability to face image.