基于边缘保护的正则化超分辨重构技术

本文研究单帧图像超分辨重构的正则化方法。利用图像分片连续、降质算子的有限支撑性和空不变性等先验信息，借鉴一元分段函数拟合的思想，构造图像超分辨重构的目标函数，设计快速有效的迭代算法，保护边缘，降低块状效应，获取高分辨图像。仿真结果表明，本文方法得到的超分辨重构图像的效果较最近邻插值法、双线性插值法的效果好。

Super-resolved Image Restoration with Edge-preserving Regularization

Super-resolved image restoration technique from a single image was studied by edge-preserving regularization. The prior knowledge of piecewise continuity, finite support and space invariability of the degraded operator were effectively exploited and utilized. The objective function was constructed and the fast iterative algorithm was designed by the thought of the piecewise univariant function-fitting and the theory of robust signal processing. The mentioned method can preserve the edge, reduce the blocky effectiveness, and reconstruct high-resolution image. Simulated result shows that the effect of super-resolved image restoration by the proposed method is better than that by the nearest-neighbor interpolation and the bilinear interpolation.