基于自适应半耦合字典学习的超分辨率图像重建

Image super-resolution reconstruction based on adaptive semi-coupled dictionary learning

在超分辨图像重建领域,如何平衡字典学习中表示系数的稀疏性和协同性对重建效果具有重要意义。针对该问题,在半耦合字典学习的超分辨重建基础上,利用核范数构建一个新的正则项,将稀疏性和协同性作为一个整体进行考虑,并用交替方向乘子法(ADMM)求解优化模型,得到了基于自适应半耦合字典学习的超分辨率图像重建算法。实验结果表明,该方法比现有的一些基于字典学习的重建方法具有更好的重建效果,其能根据字典的变化自适应地平衡稀疏性与关联性,并通过两者之间的协调产生一个最合适的系数,因此在噪声环境下具有一定的抗干扰能力。

In the field of image super-resolution,how to balance the sparsity and cooperation of the representation coefficients in the dictionary learning is of great significance for the reconstruction result.For this problem,this paper proposed an adaptive semi-coupled dictionary learning super-resolution method based on semi-coupled dictionary-learning method.The method used the kernel norm to construct a new regularization term to consider the sparsity and cooperation together,and adopted the alternating direction multiplier method(ADMM)to solve the optimal model.The experimental results show that the proposed method is more effective than some other existing dictionary-learning based reconstruction methods.It can balance the sparsity and correlation according to the variation of dictionary to produce an appropriate coefficient adaptively.Hence,it has anti-interference ability to the noisy environment.