التفاصيل البيبلوغرافية
| العنوان: |
A texture detail-oriented generative adversarial network: motion deblurring for multi-textured images. |
| المؤلفون: |
Zhang, Xiao, Chen, Ming, Zhang, Zhengqin, Lu, Shenglian |
| المصدر: |
Applied Intelligence; Feb2023, Vol. 53 Issue 3, p3255-3272, 18p |
| مصطلحات موضوعية: |
GENERATIVE adversarial networks, FEATURE extraction, COMPUTATIONAL complexity |
| الشركة/الكيان: |
GOPRO Inc. |
| مستخلص: |
The key to motion deblurring in multi-textured images is to extract more accurate feature information from the original input images. Recent studies on the design of feature extraction models have demonstrated the remarkable effectiveness of hierarchically representing multi-scale features to improve model performance. Nevertheless, these approaches generally neglect the size of the receptive field in each layer, which is essential for ensuring the full exchange of information. This paper proposes two feature extraction methods based on lightweight networks that can be flexibly plugged into networks. In addition, a novel model is formed by embedding lighter and more effective feature extraction methods (e.g., our symmetrical depthwise convolution feature extraction block (SDB) and multi-scale iterative channel feature extraction block (MIB)) into a generative adversarial network (GAN), and we name this model the texture detail-oriented GAN (TDGAN). To make the generated image closer to the target image at the visual level, we also integrate the perceptual style loss and the structural similarity loss into the generator loss function. Extensive experiments conducted on GoPro and AssIMG datasets demonstrate that the proposed model outperforms the state-of-the-art methods in terms of accuracy and computational complexity. [ABSTRACT FROM AUTHOR] |
|
Copyright of Applied Intelligence is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) |
| قاعدة البيانات: |
Complementary Index |
Full Text Finder