Underwater image enhancement using symmetrical autoencoder with synthesis datasets
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
The underwater images obtained using optical sensors, such as cameras, have low visibility and color distortion due to underwater environments. To enhance the visibility and color of underwater images, autoencoder-based convolutional neural networks are used recently; however, a basic autoencoder is not effective in color distortion and makes the result images noisy. In this study, we propose an autoencoder with skip-connection called symmetrical autoencoder (SAE) to improve the visibility and color distortion of underwater images. The proposal of this study is twofold: (i) to symmetrically add skip-connections that connect encoders to decoders throughout the network for reconstructing ability of decoders and (ii) to synthesize underwater datasets using an underwater image formation model to train the autoencoder effectively. Through the comparison with other approaches, we show that the proposed autoencoder outperforms them in PSRN, SSIM, and color difference for test datasets. In addition, the proposed autoencoder can well generalize the actual underwater images.
Abstract
The underwater images obtained using optical sensors, such as cameras, have low visibility and color distortion due to underwater environments. To enhance the visibility and color of underwater images, autoencoder-based convolutional neural networks are used recently; however, a basic autoencoder is not effective in color distortion and makes the result images noisy. In this study, we propose an autoencoder with skip-connection called symmetrical autoencoder (SAE) to improve the visibility and color distortion of underwater images. The proposal of this study is twofold: (i) to symmetrically add skip-connections that connect encoders to decoders throughout the network for reconstructing ability of decoders and (ii) to synthesize underwater datasets using an underwater image formation model to train the autoencoder effectively. Through the comparison with other approaches, we show that the proposed autoencoder outperforms them in PSRN, SSIM, and color difference for test datasets. In addition, the proposed autoencoder can well generalize the actual underwater images.
참고문헌 (11)
- G. Bianco, "A new color correction method for underwater imaging", ISPRS – International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sci- ences, 40, pp.25~32 (2015)
- O. Deperlioglu, "Underwater image enhancement with HSV and histogram equalization", 7th International Conference on Advanced Technologies (ICAT’18), pp.461~465 (2018)
- S. Anilkumar, "Algorithm for underwater cable tracking using CLAHE based enhancement", Proceedings of the International Symposium on Ocean Technology (SYMPOL), pp.129~137 (2019)
- C. O. Ancuti, "Color balance and fusion for underwater image enhancement", IEEE Transactions on Image Processing, 27(1), pp.379~393 (2018)
- K. He, "Single image haze removal using dark channel prior", IEEE Transactions on Pattern Analysis and Machine Intelligence, 33(12), pp.2341~2353 (2011)
- P. Drews Jr, "Transmission estimation in underwater single images", Proceeding of the IEEE International Conference on Computer Vision Workshops, pp.825~830 (2013)
- M. Hou, "Joint residual learning for underwater image enhancement", Proceedings of the IEEE International Conference on Image Processing, pp.4043~4047 (2018)
- L. Zhu, "Colorizing color images", Proceedings of the Electronic Imaging Symposium EI 2018 Conference on Human Vision and Electronic Imaging Science and Technology (2018)
- O. Ronneberger, "U-net : Convolutional networks for biomedical image segmentation", International Conference on Medical Image Computing and Computer-Assisted Intervention, pp.234~241 (2015)
- J. Yan, "MMP-Net : A multi-scale feature multiple parallel fusion network for single image haze removal", IEEE Access, 8, pp.25431~25441 (2020)
- X. Ding, "Jointly adversarial network to wavelength compensation and dehazing of underwater images", arXiv preprint arXiv:1907.05595, ~ (2019)