ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN Network
The preservation of image details in the defogging process is still one key challenge in the field of deep learning. The network uses the generation of confrontation loss and cyclic consistency loss to ensure that the generated defog image is similar to the original image, but it cannot retain the d...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-03-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/23/6/3294 |
| _version_ | 1797609079758127104 |
|---|---|
| author | Bingnan Yan Zhaozhao Yang Huizhu Sun Conghui Wang |
| author_facet | Bingnan Yan Zhaozhao Yang Huizhu Sun Conghui Wang |
| author_sort | Bingnan Yan |
| collection | DOAJ |
| description | The preservation of image details in the defogging process is still one key challenge in the field of deep learning. The network uses the generation of confrontation loss and cyclic consistency loss to ensure that the generated defog image is similar to the original image, but it cannot retain the details of the image. To this end, we propose a detail enhanced image CycleGAN to retain the detail information during the process of defogging. Firstly, the algorithm uses the CycleGAN network as the basic framework and combines the U-Net network’s idea with this framework to extract visual information features in different spaces of the image in multiple parallel branches, and it introduces Dep residual blocks to learn deeper feature information. Secondly, a multi-head attention mechanism is introduced in the generator to strengthen the expressive ability of features and balance the deviation produced by the same attention mechanism. Finally, experiments are carried out on the public data set D-Hazy. Compared with the CycleGAN network, the network structure of this paper improves the SSIM and PSNR of the image dehazing effect by 12.2% and 8.1% compared with the network and can retain image dehazing details. |
| first_indexed | 2024-03-11T05:55:35Z |
| format | Article |
| id | doaj.art-acc45f2ffef34da9b512a52537464787 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-11T05:55:35Z |
| publishDate | 2023-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-acc45f2ffef34da9b512a525374647872023-11-17T13:48:41ZengMDPI AGSensors1424-82202023-03-01236329410.3390/s23063294ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN NetworkBingnan Yan0Zhaozhao Yang1Huizhu Sun2Conghui Wang3School of Electronic Engineering, Xi’an Shiyou University, Xi’an 710065, ChinaSchool of Electronic Engineering, Xi’an Shiyou University, Xi’an 710065, ChinaSchool of Electronic Engineering, Xi’an Shiyou University, Xi’an 710065, ChinaSchool of Electronic Engineering, Xi’an Shiyou University, Xi’an 710065, ChinaThe preservation of image details in the defogging process is still one key challenge in the field of deep learning. The network uses the generation of confrontation loss and cyclic consistency loss to ensure that the generated defog image is similar to the original image, but it cannot retain the details of the image. To this end, we propose a detail enhanced image CycleGAN to retain the detail information during the process of defogging. Firstly, the algorithm uses the CycleGAN network as the basic framework and combines the U-Net network’s idea with this framework to extract visual information features in different spaces of the image in multiple parallel branches, and it introduces Dep residual blocks to learn deeper feature information. Secondly, a multi-head attention mechanism is introduced in the generator to strengthen the expressive ability of features and balance the deviation produced by the same attention mechanism. Finally, experiments are carried out on the public data set D-Hazy. Compared with the CycleGAN network, the network structure of this paper improves the SSIM and PSNR of the image dehazing effect by 12.2% and 8.1% compared with the network and can retain image dehazing details.https://www.mdpi.com/1424-8220/23/6/3294image defoggingCycleGANDep residual blocksmulti-head attention |
| spellingShingle | Bingnan Yan Zhaozhao Yang Huizhu Sun Conghui Wang ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN Network Sensors image defogging CycleGAN Dep residual blocks multi-head attention |
| title | ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN Network |
| title_full | ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN Network |
| title_fullStr | ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN Network |
| title_full_unstemmed | ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN Network |
| title_short | ADE-CycleGAN: A Detail Enhanced Image Dehazing CycleGAN Network |
| title_sort | ade cyclegan a detail enhanced image dehazing cyclegan network |
| topic | image defogging CycleGAN Dep residual blocks multi-head attention |
| url | https://www.mdpi.com/1424-8220/23/6/3294 |
| work_keys_str_mv | AT bingnanyan adecycleganadetailenhancedimagedehazingcyclegannetwork AT zhaozhaoyang adecycleganadetailenhancedimagedehazingcyclegannetwork AT huizhusun adecycleganadetailenhancedimagedehazingcyclegannetwork AT conghuiwang adecycleganadetailenhancedimagedehazingcyclegannetwork |