空间自适应正则化超分辨率图像重建

超分辨率图像重建是一个病态问题,在重建过程中需要正则化处理,而正则化重建会引入正则化误差及重建过程中由于病态性而引入的噪声放大误差,且这两类误差均和图像的空间局部特性有关.提出根据图像的局部空间统计特性自适应控制超分辨率图像正则化重建算法,采用图像局部统计方差来区分图像棱边区域及平滑区域,在图像的棱边区域加强图像的约束重建,而在图像的平滑区域加强正则化.实验表明该算法能有效地减小重建误差,算法的信噪比得益优于传统的正则化重建算法及总变分模型重建算法,并且对正则化参数的选择具有一定的鲁棒性.

Adaptive regularization for super-resolution image reconstruction based on local structures

Super-resolution image processing is an ill-posed problem, which needs to be regularized in the reconstruction. There are two class regularization errors in the regularized reconstruction image, which are related strongly to the local structures encountered within the image. An algorithm about super-resolution image reconstruction was proposed, which could reconstruct the super-resolution image adaptively based on the image local structures. The edge region and the smooth region were distinguished by the image local statistic variance. The observed model was reinforced in the edge region during the reconstruction, while the regularization was emphasized in the smooth region. The experiments show this algorithm is better than the traditional algorithms and the TV reconstruction algorithms, and is robust to the regularization parameter.