BTV图像超分辨率重建改进算法

Image super-resolution reconstruction based on improved bilateral total variation algorithm

在双边滤波正则化算法的基础上,结合自然图像自身结构相似性,提出一种基于BTV的改进算法。通过在代价方程中引入表达图像非局部结构相似性的正则化项,对重建图像的解空间进一步加以限制和优化,最后通过最陡下降法求得代价方程最优解,从而完成重建。实验证明,与BTV算法相比,改进后的算法不仅能很好地抑制噪声,同时也更好地保留了图像中的细节,重建图像有着比较清晰的边缘,同时该方法也说明重建过程中更多的先验知识对重建结果的重要性。

Based on the original bilateral total variation（BTV） regularization algorithm and combined with natural image itself structural similarity,a improved algorithm about BTV is proposed.The regularization term expresses image similarity of non-local structure is used in the price equation,and this method optimizates and limits the solution space of the reconstructed image.Steepest descent method is used to obtain the optimal solution of the cost equation to achieve reconstruction.Compared with the original BTV algorithm,the experiments show that the improved algorithm can not only surpress the noise,but also restore details of reconstructed image with sharper edge,and it also shows that more priori knowledge is important for the reconstruction results.