本論文研製一新穎的架構。稱為疊代低光影像增強網路，它使用暗通道先驗來增強低光源影像。我們透過觀察得知負片後之低光影像類色彩分佈似於含霧影像。因此，本論文所提出的架構遵循這個假設來恢復低光圖像。此外，我們還使用灰度世界算法來改善色彩偏移的問題。通過疊代，本架構可以得到亮度足夠的前處理影像。隨後，本論文使用自動編碼器進一步提高最終輸出影像的質量。由實驗結果可以表明，所提出的此方法可以處理各種光照條件，並且輸出效果優於現有方法。由所進行的實驗可以證明，提出之輕量化架構不僅減輕硬體設備之負擔還可以顯著提高物件偵測的性能，以便後續與高階電腦視覺任務的配合。

This paper proposes a new method called IDENet (Iterative Deep light Enhancement Network), which adopts the concept of dark channel prior to enhance low-light image. We observe that the low-light image through inverse version function is similar to haze image, which contains some pixels of very low intensities in at least one-color channel called dark channel prior. The proposed method follows this assumption to restore the low-light image. We also applied the gray world algorithm to correct color shift problem. Through iterations, we can obtain the initial version of the restored image. Then, we further improve the performance of the final output image using an auto encoder-decoder network. Experimental results show the proposed method can handle low-light images under various lighting conditions and outperforms the existing methods. Moreover, the accuracy of the object detection can be promoted by the restored image of our proposed method.

[1] Stephen M.Pizer, E. PhilipAmburn, John D.Austin, RobertCromartie, AriGeselowitz, TreyGreer, and KarelZuiderveld, “Adaptive histogram equalization and its variations,” Computer Vision, Graphics, and Image Processing, vol. 39, no. 3, pp. 355–368, 1987.
[2] A. Rizzi, C. Gatta, and D. Marini, “Color correction between gray world and white patch,” in Human Vision and Electronic Imaging VII, vol. 4662, pp. 367–375, SPIE, 2002.
[3] Y. Jiang, X. Gong, D. Liu, Y. Cheng, C. Fang, X. Shen, J. Yang, P. Zhou, and Z. Wang, “Enlightengan: Deep light enhancement without paired supervision,” IEEE Transactions on Image Processing, vol. 30, pp. 2340–2349,2021.
[4] C. Guo, C. Li, J. Guo, C. C. Loy, J. Hou, S. Kwong, and R. Cong, “Zero-reference deep curve estimation for low-light image enhancement,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 1780–1789, 2020.
[5] Jianrui Cai, Shuhang Gu, and Lei Zhang. Learning a deep single image contrast enhancer from multi-exposure image. IEEE Transactions on Image Processing, 27(4):2049–2026, 2018.
[6] Chen Wei, Wenjing Wang, Wenhan Yang, and Jiaying Liu. Deep retinex decomposition for low-light enhancement. In BMVC, 2018.
[7] S. M. Pizer, E. P. Amburn, J. D. Austin, R. Cromartie, A. Geselowitz, T. Greer, B. ter Haar Romeny, J. B. Zimmerman, and K. Zuiderveld, “Adaptive histogram equalization and its variations,” Computer vision, graphics, and image processing, vol. 39, no. 3, pp. 355–368, 1987.
[8] M. J. Shafiee, B. Chywl, F. Li, and A. Wong, “Fast yolo: A fast you only look once system for real-time embedded object detection in video,” arXiv preprint arXiv:1709.05943, 2017.