基于训练集分层的图像超分辨率重建  

Neighbor Embedding Super-resolution Reconstruction Based on the Training Set Stratification

作　　者：史云静[1] 虞涛[1] 朱秀昌[1] 

机构地区：[1]南京邮电大学江苏省图像处理与图像通信重点实验室,江苏南京210003

出　　处：《电视技术》2012年第19期18-22,63,共6页Video Engineering

基　　金：国家自然科学基金项目(61071091)

摘　　要：在基于邻域嵌入的图像超分辨率重建中,提出了一种对训练集进行分层的方法,可有效解决待重建图像块在训练集中搜索时间过长问题。同时对待重建的图像区域加以分类,对于平坦区域选择一般的双立方插值的重建方法,对于含有丰富细节的区域则采用邻域嵌入的重建方法。最后对重建图像进行IBP全局后处理,进一步提高图像质量。实验结果表明,利用本文方法重建的图像主观和客观质量都有较大的提高,且重建时间可以大大缩短。In this paper we have proposed a hierarchical method on the training set which can effectively solve the problem of long time in search in the training set when it is used in neighbor embedding super-resolution reconstruction. With this method we needn＇t care the size of training set too much, even if our training set is built very large, the search time of reconstruction will not be too long. Besides, in this paper, the reconstruction area of image will be classified, we apply the method of bicubic to reconstruct flat areas, but using our method to rebuild the image areas that are rich in details. Finally, with the use of IBP global constraint, the subjective and objective quality of reconstructed image will be further improved , but the reconstruction time can be considerably shortened .

关 键 词：超分辨率 图像重建 邻域嵌入 训练集分层 区域分类 后处理 

分 类 号：TN991.73[电子电信—信号与信息处理]