基于双正则化的图像超分辨率盲重建

Blind Image Super-resolution Reconstruction Based on Double Regularization

图像超分辨率重建是利用数字信号处理技术由一系列低分辨率观测图像得到高分辨率图像。大多数重建算法假设成像系统的模糊特性也即点扩散函数(PSF)已知,然而实际的应用环境下PSF事先不知道或部分知道。为此,将未知PSF模型化,提出基于双正则化的图像超分辨率盲重建算法,并且正则化作用的强度随重建图像局部光滑程度的变化而自适应地改变,以便能保护图像细节同时抑制平滑区域的噪声。求解过程中采用交替最小化方法估计PSF参数和高分辨率图像,并随着迭代次数的增加逐步提高每次寻优的精度以节省计算开销。实验结果表明,该算法能够比较准确地估计出PSF参数并取得较好的图像重建效果。

Image super-resolution reconstruction （SRR） refers to a signal processing approach which produces high-resolution images from observed multiple low-resolutlon images. Many image SRR algorithms assume that the blurring process, i. e. , point spread function（PSF） of the imaging system is known prior to reconstruction. However, the blurring process is not known or is known only to within a set of parameters in many practical applications. In this paper, we propose an approach for blind image SRR based on double regularization by parametrizing PSF. A space-adaptlve regularization method for image SRR is used to preserve detail at the textured regions and suppress noise in the smooth background. In the scheme, PSF parameter（s） and the high-resolution image are estimated by alternating minimization method. The demand for precision of minimizations is varied during the optimization procedure in order to reduce the computation cost. Experimental results from a synthetic image sequence show that blur parameters are approximated actually and the reconstructed image is visually pleasing.