非局部总体变分正则化的多帧图像盲反卷积算法

Multi-frame images blind deconvolution algorithm based on non local total variation regularization

针对没有完整先验知识的降质图像的多帧图像恢复问题,利用非局部总体变分正则化方法,提出了多帧图像盲反卷积问题求解的有效数值算法。该算法既能保持重建图像的边缘与纹理结构又能抑制相关噪声,而且能同时重建原始图像和相关的点扩展函数PSF。实验结果表明提出的方法具有明显的优越性。

In order to solve multi-frame images restoration problem without complete prior knowledge, a non total varia-tion regularization method is applied to reconstruct original image and the associated Point Spread Function（PSF）, a effi-cient numerical scheme is proposed to solve the variational problem. The proposed numerical scheme can preserve sharp edges and fine structures, textures of images while alleviating the concerns of noise amplification. Numerical experimen-tal results show the excellent performance of the proposed method.