增益映射耦合局部正则化的图像重构算法

Super resolution image reconstruction algorithm based on gain map and local regularization

针对当前的图像重构方法在对多帧超分辨率图像复原时,存在明显的模糊效应与振铃效应的不足,提出增益映射控制耦合局部正则化的图像重构算法。首先,通过对低分辨率图像中亚像素的移动进行分析,构建高低分辨率图像的成像模型,再对超分辨率图像进行估值,将重构问题转化为一个不稳定的线性方程式组;然后,构造正则化算子,联合改进的代数重建法求其稳定值;最后,采用基于局部自适应正则化的增益可控方法建立增益映射,完成超分辨率图像重构。仿真结果表明,与当前图像重构算法相比,在修复低分辨率图像时,该机制拥有更好的重构效果,有效降低了模糊效应与振铃效应。

Owing to these defects such as blurring effect and ringing effect, caused by the current image reconstruction algorithm for the ill- posed multi- frame super- resolution image, this paper proposes an effective mechanism based on local adaptive regulariza-tion for image reconstruction. First, by analysis of the sub- pixel shifts in low- resolution image, the high and low resolution forma-tion model is constructed, and by estimate an HR image, the reconstruction problems are transformed into an unstable linear equa-tions. Then, a regularization operator is constructed, combined modified algebraic reconstruction technique for the solution of the lin-ear equations. Finally, the gain control method based on local adaptive regularization is used to realize construction of gain map,and complete the SR image reconstruction. The simulation results show that comparing with current image reconstruction algorithm,the reconstruction performance of this mechanism is better, which effectively reduces the blurring effect and ringing effect under the condition of LR image.