近似稀疏正则化的红外图像超分辨率重建

针对红外图像分辨率低、受噪声影响严重等问题，引入近似稀疏正则化和K-奇异值分解（K-SVD）法，提出了基于近似稀疏表示模型的红外图像超分辨率重建方法。考虑到红外图像受到噪声污染，首先建立了稳健近似稀疏表示模型。针对已有字典训练方法时间消耗巨大问题，在假定低分辨率图像空间和高分辨率图像空间具有相似流形的前提下，联合近似稀疏表示模型和K-SVD方法，提出近似稀疏约束的基于K-SVD的高低分辨率字典对学习算法。最后，通过高分辨字典和对应的红外图像群稀疏表示系数重建得到高分辨率的红外图像。为了验证算法的性能，对提出的算法与稀疏性正则化的图像超分辨模型（SRSR）和Zeyde算法进行了实验比较。结果表明，本文方法能够较好地减少红外图像中的噪声，同时获得更好的超分辨率重建效果。

Super-resolution reconstruction of approximate sparsity regularized infrared images

For the problems of low-resolution and serious effect from noises of infrared images, an approximate sparsity regularized infrared image super-resolution reconstruction algorithm (ASSR) based on K-SVD (Singular Value Decomposition) was proposed. In consideration of the noise effect from infrared images, an approximate sparsity representation model was first established. On the assumption that the low and high resolution image spaces hold a similar manifold, an approximate sparsity regularized K-SVD based dictionary learning method was proposed with approximate sparsity model and K-SVD method to solve the time-consuming problem of existing dictionary training process. Finally, the high-resolution infrared images were recovered by the high-resolution dictionary and the corresponding low-resolution group sparse coefficients. To verify the performance of the algorithm proposed, it was compared with those of the Sparsity Regularized Super-Resolution Reconstruction (SRSR) and Zeyde algorithm. Experimental results show that the proposed method can reduce the noises of infrared images, and can obtain excellent performance in super-resolution reconstruction.