摘要
针对传统超分辨率图像重建算法速度慢的缺点,提出了一种基于自适应各向异性正则化的快速超分辨率图像重建算法。本文算法兼顾重建图像质量的同时,提升了图形的重建速度。基于传统迭代算法,本文算法通过优化约束条件,大量剔除了冗余过程,弥补了传统算法的不足;同时引入一种具有自适应能力的各向异性平滑项,可以适应各种复杂的运动模型。另外,提出以图像的峰值信噪比(PSNR)为标准,作为重建迭代的截止条件。运用本文算法对序列低分辨率图像进行重建,证明了本文算法可以更快实现超分辨率图像重建。
A fast super-resolution reconstruction algorithm based on adaptive anisotropic regularization is proposed to overcome the low speed of traditional super-resolution reconstruction algorithms. This algo- rithm can improve the speed of image reconstruction,while the quality of the image is also reconstructed very well. Based on the traditional iterative model,the constraint condition is optimized,and the redun- dant processes are deleted to avoid the disadvantages of traditional methods. A adaptive anisotropic smoothing term is proposed ot preserve the sharper edges effectively. In addition,the cut-off condition of reconstructed iteration is proposed based on the peak signal-to-noise ratio （PSNR） of the image. The al- gorithm is applied to reconstruct the sequence low-resolution image, and it is proved that the algorithm can realize the super-resolution reconstruction more quickly.
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2017年第7期788-793,共6页
Journal of Optoelectronics·Laser
基金
国家自然科学基金(61171164)资助项目
关键词
超分辨率
图像重建
鲁棒性
去模糊
各向异性
super-resolution
image reconstruction
robustness
deblurring
anisotropic
