基於人眼視覺具有去噪及色彩校正之低亮度影像增強

Retinex-based Low-Light Image Enhancement with Denoising and Color Correction

林玟君(Wen-Chun Lin)

指導教授 ： 貝蘇章

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2021年)

https://doi.org/10.6342/NTU202100941

摘要

在低亮度的環境下，所拍攝的影像通常會面臨許多問題，像是雜訊較大、光源照度不均勻、曝光度不足導致暗處細節與色彩能見度低，而本篇論文主要建構一個低亮度影像增強系統來解決上述問題，並對此系統內部的色彩校正演算法有詳盡的介紹。
於我所提出的低亮度影像增強系統中，分成三個部分進行處理。首先是對低亮度影像做降噪的前處理，有鑑於影像增強的過程中，會導致雜訊跟著被放大與強化，因此我將去雜訊的步驟放置在整個系統的最前線。接著是對降噪後的低亮度影像做亮度的增強，此處以Retinex理論為基礎，將影像分解為光照分量與反射分量，並結合Camera Response Function來估測實際光照成分，透過提高其低光照區的亮度，進而獲取亮度強化之結果。最後是做色彩校正的後處理，由於提高亮度後的影像色彩會有所偏離，甚至因局部強力人造光源造成色偏現象發生，因此我也提出了不同方案去做處理，一種是修正色彩至與原低亮度影像色彩相近，另一種則是透過統計特性，融合白平衡與強力光源，以符合人眼色彩恆常性。

關鍵字

低亮度影像增強 ； 降低雜訊 ； 相機響應函數估測 ； 光照估測 ； 色彩校正 ； 色彩恆常性 ； 色彩轉換

並列摘要

Images usually suffer from large noise, non-uniform illumination and low visibility due to insufficient exposure when captured in low-light conditions. In this thesis, we introduce a low-light image enhancement system to overcome these problems, and also emphasize a color correction algorithm applied in this system.
There are three parts in the low-light image enhancement system. The first one is to do the pre-processing of noise reduction. Since the process of image enhancement may increase the noise, the noise reduction step is placed at the front of the system. The second one is to enhance the illumination of the image. Based on Retinex theory, the image can be decomposed into two components, illumination and reflectance. The illumination can be estimated precisely by using the camera response function, which transfers irradiance to image intensities. After the region of the low illumination is lightened, the result of illumination enhancement can be acquired. The last one is to do the post-processing of color correction. Here, we propose two different methods. One is preserving the original color information. The other is combining white balance and the information of local light sources to retain color constancy by using statistical properties.

並列關鍵字

Low-Light Image Enhancement ； Noise Reduction ； Camera Response Function Estimation ； Illumination Estimation ； Color Correction ； Color Constancy ； Color Transfer

參考文獻

[1] C. H. Ooi and N. A. M. Isa, "Quadrants dynamic histogram equalization for contrast enhancement," IEEE Transactions on Consumer Electronics, vol. 56, no. 4, pp. 2552-2559, 2010.

[2] K. Singh, R. Kapoor, and S. K. Sinha, "Enhancement of low exposure images via recursive histogram equalization algorithms," Optik, vol. 126, no. 20, pp. 2619-2625, 2015.

[3] X. Guo, Y. Li, and H. Ling, "LIME: Low-light image enhancement via illumination map estimation," IEEE Transactions on image processing, vol. 26, no. 2, pp. 982-993, 2016.

[4] X. Fu, D. Zeng, Y. Huang, X.-P. Zhang, and X. Ding, "A weighted variational model for simultaneous reflectance and illumination estimation," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016, pp. 2782-2790.

[5] S.-Y. Yu and H. Zhu, "Low-illumination image enhancement algorithm based on a physical lighting model," IEEE Transactions on Circuits and Systems for Video Technology, vol. 29, no. 1, pp. 28-37, 2017.