过完备字典稀疏表示的云图超分辨率算法

Nephogram super-resolution algorithm using over-complete dictionary via sparse representation

提出一种基于过完备字典稀疏表示的云图超分辨率算法。首先,联合训练针对低分辨率与高分辨率云图块的两个字典Dl和Dh,保证对应的低分辨率与高分辨率云图块关于各自的字典具有相似的稀疏表示;其次,通过求解优化问题,获得待处理云图每个低分辨率云图块关于Dl的稀疏表示,并将表示系数用于Dh以生成对应的高分辨率云图块;最后,运用最速下降算法,得到满足重构约束的高分辨率云图。红外与可见光云图的数值实验验证了本文算法的有效性,表明本文算法在视觉效果及PSNR指标上均优于插值方法。

Motivated by the fact that image patch can be sparse represented using a suitable over-complete dictionary, a nepho- gram super-resolution algorithm via sparse representation using over-complete dictionary is presented. During the experiment two dictionaries DI and Dh for the low-resolution and high-resolution nephogram patches were trained jointly in order to guar- antee that the low-resolution and high-resolution patch pair possesses similar sparse representations as to their own dictionary. Through solving optimization problem, the sparse representation for each low-resolution nephogram patch with respect to DI was obtained, and the representation coefficients were applied to Dh in order to generate the corresponding high-resolution nephogram patch. At the end of experiment the high-resolution nephogram which satisfies the reconstruction constraint was achieved by us- ing gradient descent algorithm. Numerical experiments for infrared and visual nephogram demonstrate the effectiveness of the proposed algorithm. Moreover, the proposed algorithm outperforms interpolation based methods in terms of visual quality and the Peak Signal to Noise Ratio （PSNR）.