A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous Losses
We propose a novel generative adversarial network (GAN)-based image denoising method that utilizes heterogeneous losses. In order to improve the restoration quality of the structural information of the generator, the heterogeneous losses, including the structural loss in addition to the conventional...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-02-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/21/4/1191 |
| _version_ | 1827601891769450496 |
|---|---|
| author | Sung In Cho Jae Hyeon Park Suk-Ju Kang |
| author_facet | Sung In Cho Jae Hyeon Park Suk-Ju Kang |
| author_sort | Sung In Cho |
| collection | DOAJ |
| description | We propose a novel generative adversarial network (GAN)-based image denoising method that utilizes heterogeneous losses. In order to improve the restoration quality of the structural information of the generator, the heterogeneous losses, including the structural loss in addition to the conventional mean squared error (MSE)-based loss, are used to train the generator. To maximize the improvements brought on by the heterogeneous losses, the strength of the structural loss is adaptively adjusted by the discriminator for each input patch. In addition, a depth wise separable convolution-based module that utilizes the dilated convolution and symmetric skip connection is used for the proposed GAN so as to reduce the computational complexity while providing improved denoising quality compared to the convolutional neural network (CNN) denoiser. The experiments showed that the proposed method improved visual information fidelity and feature similarity index values by up to 0.027 and 0.008, respectively, compared to the existing CNN denoiser. |
| first_indexed | 2024-03-09T05:06:48Z |
| format | Article |
| id | doaj.art-3d07133248c04e11867f2e4c2d7429aa |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-09T05:06:48Z |
| publishDate | 2021-02-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-3d07133248c04e11867f2e4c2d7429aa2023-12-03T12:53:25ZengMDPI AGSensors1424-82202021-02-01214119110.3390/s21041191A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous LossesSung In Cho0Jae Hyeon Park1Suk-Ju Kang2Department of Multimedia Engineering, Dongguk University, Seoul 04620, KoreaDepartment of Multimedia Engineering, Dongguk University, Seoul 04620, KoreaDepartment of Electrical Engineering, Sogang University, Seoul 121-742, KoreaWe propose a novel generative adversarial network (GAN)-based image denoising method that utilizes heterogeneous losses. In order to improve the restoration quality of the structural information of the generator, the heterogeneous losses, including the structural loss in addition to the conventional mean squared error (MSE)-based loss, are used to train the generator. To maximize the improvements brought on by the heterogeneous losses, the strength of the structural loss is adaptively adjusted by the discriminator for each input patch. In addition, a depth wise separable convolution-based module that utilizes the dilated convolution and symmetric skip connection is used for the proposed GAN so as to reduce the computational complexity while providing improved denoising quality compared to the convolutional neural network (CNN) denoiser. The experiments showed that the proposed method improved visual information fidelity and feature similarity index values by up to 0.027 and 0.008, respectively, compared to the existing CNN denoiser.https://www.mdpi.com/1424-8220/21/4/1191image denoisingconvolutional neural networkgenerative adversarial networkimage restorationstructural loss |
| spellingShingle | Sung In Cho Jae Hyeon Park Suk-Ju Kang A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous Losses Sensors image denoising convolutional neural network generative adversarial network image restoration structural loss |
| title | A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous Losses |
| title_full | A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous Losses |
| title_fullStr | A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous Losses |
| title_full_unstemmed | A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous Losses |
| title_short | A Generative Adversarial Network-Based Image Denoiser Controlling Heterogeneous Losses |
| title_sort | generative adversarial network based image denoiser controlling heterogeneous losses |
| topic | image denoising convolutional neural network generative adversarial network image restoration structural loss |
| url | https://www.mdpi.com/1424-8220/21/4/1191 |
| work_keys_str_mv | AT sungincho agenerativeadversarialnetworkbasedimagedenoisercontrollingheterogeneouslosses AT jaehyeonpark agenerativeadversarialnetworkbasedimagedenoisercontrollingheterogeneouslosses AT sukjukang agenerativeadversarialnetworkbasedimagedenoisercontrollingheterogeneouslosses AT sungincho generativeadversarialnetworkbasedimagedenoisercontrollingheterogeneouslosses AT jaehyeonpark generativeadversarialnetworkbasedimagedenoisercontrollingheterogeneouslosses AT sukjukang generativeadversarialnetworkbasedimagedenoisercontrollingheterogeneouslosses |