改进的单帧图像超分辨率算法研究  

Improved super-resolution algorithm of single-frame image

作　　者：田萌[1,2] 修春波[1,2] 卢少磊[1,2] 

机构地区：[1]天津工业大学电工电能新技术天津市重点实验室,天津300387 [2]天津工业大学电气工程与自动化学院,天津300387

出　　处：《计算机工程与应用》2014年第8期150-153,共4页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61203302;No.61078041)

摘　　要：针对牛顿插值算法信息利用率不高的缺点,提出了一种改进的单帧图像超分辨率算法。利用源图像中的像素点信息,同时从多个方向计算牛顿插值结果,并根据源图像中各像素点的相关性通过融合计算获得超分辨率图像的插值结果。该方法既提高了源图像中信息的利用率,又减小了插值误差的累积。仿真实验结果表明,利用改进方法所获得的超分辨率图像更细腻清晰,尤其图像边缘区域所包含的大量细节信息可得到有效恢复。改进算法所获得的超分辨率图像的峰值信噪比、均方误差以及视觉信息保真度等评价指标均优于传统方法。In order to improve the Newton interpolation algorithm, a new super-resolution algorithm of single-frame image is proposed. The Newton interpolation results are computed from multi-directions by the pixels in the original image. The reconstruction results of super-resolution image are obtained by the fusion of interpolation results according to the correla-tion of the pixels in the original image. The improved method can not only enhance the utilization rate of the information contained in the original image, but also decrease accumulation error of interpolation. Simulation results show that the super-resolution image obtained by the improved algorithm is fine and distinct. A lot of detail information in marginal regions can be recovered validly. The method has better key performance indicators, such as peak signal-to-noise ratio, mean square error and visual information fidelity, than conventional methods.

关 键 词：超分辨率 牛顿插值 相关性 单帧图像 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]