An Image Quality Evaluation Method Based on Joint Deep Learning
Lists
Only you see this
Abstract
The image quality plays a very important role in image processing. In this paper, we propose an image quality evaluation method based on joint deep learning (JDL). Specifically, deep belief networks (DBNs) and convolutional neural network (CNNs) will be used together. Both the features extracted by human and features extracted by machine will be used to evaluate the quality of the images. For DBNs framework, the image features in both spatial domain and transform domain will be extracted. Then, it will be used to efficiently calculate and finally obtain image quality, $$Q_{1}$$ . For CNNs, the framework will calculate the image quality without features extracted by machine, which can be defined as $$Q_{2}$$ . At last, joint framework will give the final assessment result. Experiments show that our method is very consistent with the actual subjective assessment result.
BibTeX
@inproceedings{yang2017image,
author = {Jiachen Yang and Bin Jiang and Yinghao Zhu and Chunqi Ji and Wen Lu},
title = {{An Image Quality Evaluation Method Based on Joint Deep Learning}},
booktitle = {ICONIP},
year = {2017},
doi = {10.1007/978-3-319-70087-8\_68},
}Loading related content...