Treffer: Transmission-guided multi-feature fusion Dehaze network.

Image dehazing is an important direction of low-level visual tasks, and its quality and efficiency directly affect the quality of high-level visual tasks. Therefore, how to quickly and efficiently process hazy images with different thicknesses of fog has become the focus of research. This paper presents a multi-feature fusion embedded image dehazing network based on transmission guidance. Firstly, we propose a transmission graph-guided feature fusion enhanced coding network, which can combine different weight information and show better flexibility for different dehazing information. At the same time, in order to keep more detailed information in the reconstructed image, we propose a decoder network embedded with Mix module, which can not only keep shallow information, but also allow the network to learn the weights of different depth information spontaneously and re-fit the dehazing features. The comparative experiments on RESIDE and Haze4K datasets verify the efficiency and high quality of our algorithm. A series of ablation experiments show that Multi-weight attention feature fusion module (WA) module and Mix module can effectively improve the model performance. The code is released in https://doi.org/10.5281/zenodo.10836919. [ABSTRACT FROM AUTHOR]

Copyright of Visual Computer is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. 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.)