摘要
为了进一步提高基于学习的超分辨率图像重建质量,考虑到极限学习机(ELM)具有学习速度快和良好数据预测与分析能力,提出了1种基于极限学习机的图像超分辨率重建方法.在图像稀疏思想下,将高分辨率图像中的高频细节信息作为原子构建冗余字典.具体是提取训练图像的高频信息,采用改进的K-SVD算法对高低分辨率图像进行字典学习,构建对应的特征字典作为极限学习机的输入训练网络参数,建立超分辨率重建模型.最后仿真实验结果表明,所提算法能取得比对比算法更好的实验数据.
In order to further improve the quality of the learning-based super-resolution image reconstruction, considering the extreme learning machine (ELM) with fast learning speed and good data prediction and analysis, this paper propos- es image super resolution reconstruction based on the extreme learning machine. Under the idea of image sparse, high-fre- quency details is used as atomic to construct redundant dictionary. Specifically, high frequency information of the training image is extracted. The improved K-SVD algorithm is used to carry out dictionary learning on high and low resolution im- ages. The corresponding feature dictionary is constructed as the input to train network parameter. Super-resolution reconstruction model is established. Finally the simulation results show that the proposed algorithm can obtain better experimental data than the comparative algorithm.
出处
《河北工业大学学报》
CAS
2017年第2期11-16,共6页
Journal of Hebei University of Technology
基金
天津市自然科学基金(14JCZDJC32600)
关键词
极限学习机
字典学习
超分辨率
高频信息
extreme learning machine
dictionary learning
super resolution
high frequency information