基于序列图像的超分辨率重建  

Super-resolution Reconstruction Based on Image Sequences

作　　者：樊超[1] 孙宁宁[1] 夏旭[1] 

机构地区：[1]河南工业大学信息科学与工程学院,河南郑州450001

出　　处：《红外技术》2010年第5期279-282,共4页Infrared Technology

基　　金：河南工业大学博士基金项目;编号:2007BS033

摘　　要：提出一种从序列低分辨率欠采样图像中重建出一幅高分辨率图像的重建方法。该方法基于MAP算法,通过利用梯度投影的方法对重建结果不断进行迭代优化得到最终的理想高分辨率影像。实验结果证明,该方法有效,它不仅能够非常容易的引入各种先验条件进行约束从而提高解的收敛速度,并且能够得到质量较好的高分辨率重建图像。在取得相同质量图像的前提下,该算法所需时间明显降低。A new super-resolution algorithm is proposed to the problem of reconstructing a high-resolution image from several low-resolution images that have been sub-sampled in this paper. In order to attain a high-resolution image, the algorithm is based on the MAP algorithm, solving the optimization by proposed iteration steps with using the gradient projection method. The results of the experiments indicate that the proposed algorithm has considerable effectiveness in that it can not only be very easy to utilize the priori conditions as constraints to improve the convergence rate, but also can get the high resolution reconstruction image expectedly. Simulation results demonstrate that the method can realize the super-resolution image with little time.

关 键 词：超分辨率 图像重建 最大后验概率 梯度投影 

分 类 号：TP911.73[自动化与计算机技术]