CAJ | 학술논문

将低分辨率图像重建成高分辨率图像是图像处理领域中的一个重要课题。Yang提出一种基于联合字典学习的图像超分辨率重建算法，其算法样本选取与字典训练方法较为复杂。提出一种基于MOD字典学习的图像超分辨率重建新算法，首先采用少量的训练样本代替Yang的大量训练样本，然后使用MOD字典学习算法代替Yang的FFS字典学习算法，最后利用字典对图像进行稀疏表示与重建。实验结果表明，所提出的算法速度较快，并且重建图像的质量较高。
장저분변솔도상중건성고분변솔도상시도상처리영역중적일개중요과제。Yang제출일충기우연합자전학습적도상초분변솔중건산법，기산법양본선취여자전훈련방법교위복잡。제출일충기우MOD자전학습적도상초분변솔중건신산법，수선채용소량적훈련양본대체Yang적대량훈련양본，연후사용MOD자전학습산법대체Yang적FFS자전학습산법，최후이용자전대도상진행희소표시여중건。실험결과표명，소제출적산법속도교쾌，병차중건도상적질량교고。
It is an important topic to reconstruct a high resolution image from a low resolution image. Yang proposed an image super-resolution reconstruction algorithm based on the joint dictionary-learning, which needs large samples, and dictionary training methods are complicated. In this paper, a new algorithm of image super-resolution reconstruction based on MOD dictionary-learning is proposed, a small amount of training samples is firstly used to replace large numbers of training samples of Yang?s, then the MOD dictionary-learning algorithm is used instead of Yang?s FFS dictionary-learning algorithm, at last, the resulted dictionary is applied to the image sparse representation and super-resolution reconstruction. The experimental results show that the image reconstruction speed is improved greatly with better reconstruction quality.