基於多尺度融合的水下影像修復技術

Underwater Image Restoration Based on Multi-scale Fusion

李俊棋(Jun-Qi Li)

指導教授 ： 張恆華

國立臺灣大學/工學院/工程科學及海洋工程學研究所/碩士(2018年)

https://doi.org/10.6342/NTU201803048

摘要

在水下技術研究領域裡，水下載具通常配備視覺系統，可捕獲感興趣的生物或礦物的各種圖像並監測環境條件。遺憾的是，捕獲的水下圖像通常具有嚴重的顏色失真和較差的可見度問題，這是因為水下影像受到水裡濃密混濁介質與普遍存在的懸浮微粒影響，使得光在水中傳遞時產生衰減、吸收及散射問題，導致水下影像的對比度嚴重下降；此外，根據波長的性質，衰減的幅度也會有所不同。本研究中，我們主要根據簡化的水下光學模型來有效率地修復水下影像。首先，我們使用兩種不同演算法所求得的傳遞率與背景光，基於融合原理和特定的權重，得到融合後的傳遞率與背景光，如此一來，其能見度將會以物體與相機之間的相對距離獲得適當的補償。接者，經由分析點擴散方程式與前向散射關係，使用低通濾波來對水下影像去捲積。最後，我們均化各個色彩頻道亮度平均值以平衡顏色。實驗結果顯示，本研究所提方法比起許多現有先進的技術，可以獲得更良好的修復品質和視覺質量。

關鍵字

水下影像 ； 影像修復 ； 圖像融合顯著性圖。

並列摘要

In the field of undersea related research, underwater vehicles usually carry a visual system that captures various images of interested creatures, minerals and monitors environmental conditions. Unfortunately, the captured images often have serious color distortion and poor visibility problems. This is because that underwater images are usually affected by the turbid water medium and floating particles existed in the water. Three different problems of attenuation, absorption, and scattering happen while light propagates in the water. These phenomena cause low contrast in underwater images. Furthermore, the quantity of attenuation is associated with the wavelength of light . In this thesis, we simplifies the optical model and proposes an effective algorithm to recover underwater images. First, we compute the background light and transmission using two different algorithms. Based on the fusion principle and specific weights in between, we can obtain better the background light and transmission after fusion. The visibility of scene is compensated by the object-camera distance to recover the color of the background and objects. Subsequently, by realizing the physical property of the point spread function, we adopted a low-pass filter to deblur the image by deconvolution. Finally, we equalize the color mean in each channel to balance the color. Comparing with many existing methods, our method demonstrated better restoration results and visual quality.

並列關鍵字

underwater image ； image restoration ； image fusion ； saliency map

參考文獻

[1] G. Griffiths, Technology and applications of autonomous underwater vehicles. CRC Press, 2002.

[2] D. F. Coleman, J. B. Newman, and R. D. Ballard, "Design and implementation of advanced underwater imaging systems for deep sea marine archaeological surveys," in Oceans 2000 MTS/IEEE Conference and Exhibition, 2000, vol. 1, pp. 661-665: IEEE.

[3] Y. Y. Schechner and N. Karpel, "Clear underwater vision," in Computer Vision and Pattern Recognition, 2004. CVPR 2004. Proceedings of the 2004 IEEE Computer Society Conference on, 2003, vol. 1, pp. I-I: IEEE.

[4] R. Schettini and S. Corchs, "Underwater image processing: state of the art of restoration and image enhancement methods," EURASIP Journal on Advances in Signal Processing, vol. 2010, p. 14, 2010.

[5] A. Arnold-Bos, J.-P. Malkasse, and G. Kervern, "A preprocessing framework for automatic underwater images denoising," ed, 2005.