摘要
针对不理想的配准结果会导致超分辨率重建失败的问题,提出了一种基于加速健壮特征(SURF)匹配和凸集投影(POCS)的超分辨率重建算法。该算法首先采用SURF算法进行连续帧图像的配准,估计图像序列的运动位移;然后根据运动估计结果,在POCS理论框架下进行图像重建。实验结果表明,该方法能够较明显地改善图像的视觉效果,获得较丰富的细节信息,且具有较好的噪声抑制能力。
To solve the problem that imprecise registration will lead to the failure of the super-resolution reconstruction, an algorithm of super-resolution sequential image reconstruction based on Speeded Up Robust Feature (SURF) registration method and Projection Onto Convex Set (POCS) was presented. Firstly, SURF algorithm was utilized to register sequential images and estimate translational motion among various frames. Then, image reconstruction was carried out in the POCS framework. Experimental results have shown this method can obviously improve the visual effect of images and obtain rich details, also it can effectively suppress noise.
出处
《计算机应用》
CSCD
北大核心
2012年第A02期159-161,共3页
journal of Computer Applications
基金
上海市教委重点学科建设项目(J51801)
上海第二工业大学校科研基金资助项目(A20XK11X014)
上海市教育委员会科研创新项目(13YZ129)
关键词
超分辨率
凸集投影
积分图
加速稳健特征
特征匹配
super-resolution
Projection Onto Convex Set (POCS)
integral-image
Speeded Up Robust Feature (SURF)
feature matching