新型混合字典学习算法改进图像超分辨率的研究
目前机器学习算法大都采用单层字典的学习训练设计，现从改进视觉效果、分辨率2个角度，设计了2层字典混合学习算法。采用经典的半耦合字典学习（SCDL）和模糊模型处理结合作为第1层字典学习，第2层则是针对第1层的残余图像进行重构，结合K值聚类以及K—SVD算法设计了第2层字典的训练过程。与经典SCSR、SCDL算法对比实验结果表明：改进算法的峰值信噪比与其他2种算法有了4％左右的提高，提高值在1dB以上，表明了算法能够一定程度的提高重构图像的分辨率；算法的对比视觉效果看出，改进的算法能够明显改进重构质量，实现了图像视觉效果的改善。由于算法并不是以牺牲算法运算时间、速度为代价，这样其研究结果对于机器学...
Saved in:
| Published in | 实验室研究与探索 Vol. 36; no. 11; pp. 118 - 121 |
|---|---|
| Main Author | |
| Format | Journal Article |
| Language | Chinese |
| Published |
广东开放大学,广州,510091%兰州理工大学经济与管理学院,兰州,730050
2017
|
| Subjects | |
| Online Access | Get full text |
| ISSN | 1006-7167 |
Cover
| Summary: | 目前机器学习算法大都采用单层字典的学习训练设计，现从改进视觉效果、分辨率2个角度，设计了2层字典混合学习算法。采用经典的半耦合字典学习（SCDL）和模糊模型处理结合作为第1层字典学习，第2层则是针对第1层的残余图像进行重构，结合K值聚类以及K—SVD算法设计了第2层字典的训练过程。与经典SCSR、SCDL算法对比实验结果表明：改进算法的峰值信噪比与其他2种算法有了4％左右的提高，提高值在1dB以上，表明了算法能够一定程度的提高重构图像的分辨率；算法的对比视觉效果看出，改进的算法能够明显改进重构质量，实现了图像视觉效果的改善。由于算法并不是以牺牲算法运算时间、速度为代价，这样其研究结果对于机器学习在图像领域的进一步推广与发展具有一定的参考价值。 |
|---|---|
| Bibliography: | LI Ke1 , SUN Jinling2（1. The Open University of Guangdong, Guangzhou, 510091, China; 2. School of Economics and Management,Lanzhou University of Technology, Lanzhou 730050, China） At present the machine learning algorithms mostly adopt single dictionary learning training design. From the perspectives of improving simultaneously resolution and visual effect, a 2-layer dictionary hybrid learning algorithm is designed. The classic coupling half dictionary learning （SCDL） is combined with fuzzy model to use as the first level dictionary, learning, the second layer is used to reconstruct the residual image of the first layer. The second level is trained by the K value clustering and the K-SVD algorithm. Compared with classical SCSR, SCDL algorithms, experimental results show that the improved algorithm can improve the peak signal-to-noise ratio about 4% , and tile value is more than 1 dB. It illustrates that the algorithm can improve the resolution of the reconstructed image and visual effect. Because the algorithm |
| ISSN: | 1006-7167 |