| Title |
Contrast Enhancement Based on Encoder-decoder Network Using Multi-Scale Feature Map |
| Authors |
변상현(Sanghyun Byun) ; 임헌성(Heunseung Lim) ; 유수환(Soohwan Yu) ; 백준기(Joonki Paik) |
| DOI |
https://doi.org/10.5573/ieie.2020.57.5.74 |
| Keywords |
Contrast enhancement; Deep-learning; Multi-scale feature map; Encoder-decoder network |
| Abstract |
The paper presents a contrast image enhancement algorithm based on an encoder-decoder network. The proposed method consists of three steps: i) multi-scale feature map generation from an input low-contrast image, ii) contrast enhancement based on the encoder-decoder network, and iii) detail reconstruction network. The proposed method effectively prevents color distortion by combining the multi-scale feature maps generated from the input image while enhancing the contrast using the encoder-decoder network. Besides, the detail reconstruction network restores the edges and textures lost in the compression layers of the encoder-decoder network. Experimental results show that the proposed method can provide the contrast-enhanced result without brightness saturation and color distortion than existing contrast enhancement methods. |