基于流形学习和梯度约束的图像超分辨率重建  

Super-Resolution Image Reconstruction Based on Manifold Learning and Gradient Constraint

作　　者：廖秀秀[1] 韩国强[1] 沃焱[1] 黄汉铨[1] 李展[2] 

机构地区：[1]华南理工大学计算机科学与工程学院,广东广州510006 [2]暨南大学计算机科学系,广东广州510632

出　　处：《华南理工大学学报（自然科学版）》2012年第4期8-15,共8页Journal of South China University of Technology(Natural Science Edition)

基　　金：NSFC-广东省联合基金资助项目(U1035004);国家自然科学基金青年科学基金资助项目(61003270);国家自然科学基金面上项目(61070090);广东省工业攻关科技计划项目(2009B030803004);华南理工大学中央高校基本科研业务费专项资金重点资助项目(2012ZZ0066);广东省重大科技专项项目(2010A080402005);广东省自然科学基金博士启动项目(10452840301004638)

摘　　要：将改进的基于流形学习的超分辨率重建与基于梯度约束的正则化重建结合起来,提出一种新的单帧图像超分辨率重建算法.该算法首先针对基于流形学习的超分辨率重建,提出新的特征提取方法,联合归一化亮度与平稳小波变换细节子带系数两个特征矢量,提高重建性能;然后将学习得到的高分辨率图像作为初始估计,将其梯度作为目标梯度域,进行基于梯度约束的正则化重建,得到最终的高分辨率图像.与现有的一些算法相比,文中算法无论在视觉效果还是客观评价上都具有较好的重建性能.Proposed in this paper is a novel super-resolution reconstruction algorithm of single-frame images, which integrates the improved super-resolution reconstruction based on manifold learning with the regularized reconstruc- tion based on gradient constraint. In this algorithm, a new feature extraction method, which combines the two fea- ture vectors of the normalized luminance and the detail sub-band coefficient of stationary wavelet transform, is put forward for the super-resolution reconstruction based on manifold learning, and is used to improve the reconstruction performance. Then, a regularized reconstruction based on gradient constraint is implemented to obtain the final high-resolution image, with the learned high-resolution image and its gradient respectively as the initial estimate and the target gradient field. As compared with some existing algorithms, the proposed algorithm is of better reconstruc- tion performance in terms of both visual effect and objective evaluation.

关 键 词：图像处理 超分辨率重建 流形学习 梯度约束 正则化重建 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]