用於影像除霧的新式介質圖

A New Transmission Map for Image Dehazing

廖俊嘉(Jyun-Jia Liao)

指導教授 ： 黃士嘉

國立臺北科技大學/電資學院/電腦與通訊研究所/碩士(2014年)

摘要

由於霧氣天氣環境的大氣粒子吸收與散射的影響，導致戶外場景影像的視覺能見度被降低。而這些被霧化的影像可能會嚴重地劣化智慧型運輸系統的效能，如交通狀態監測系統、車輛監控系統以及障礙物偵測系統等。近年來，霧氣移除技術應用於這些電腦視覺系統上的議題，已逐漸被受到注意與重視，主要是為了能提高霧氣影像的能見度，以達到提升這些電腦視覺系統的效能與可靠性。然而，對於在真實場景的霧氣影像中，要有效且精確地評估霧氣濃度於介質圖上是相當困難的。因此，本文提出一個有效於真實霧氣影像評估介質圖的能見度修復方法。我們方法主要由兩個模組所構成，分別為霧氣濃度評估模組與能見度修復模組。其中霧氣濃度評估模組透過我們所提出的適應性雙伽瑪校正技術，能有效且精確地評估霧氣濃度於介質圖中，進而透過此介質圖資訊與能見度修復模組來有效修復於霧氣天氣下所擷取的真實影像。透過視覺分析的實驗證明，我們所提出的方法比其他先進方法更能達到優異的修復效力，以及獲取更清晰的視覺能見度。

關鍵字

暗通道先驗 ； 霧氣移除

並列摘要

Visibility of the captured outdoor images in inclement weathers, such as haze, fog and mist, usually is degraded due to the effect of absorption and scattering caused by the atmospheric particles. Such images may significantly contaminate the performance qualities of the intelligent transportation systems relying on visual feature extraction, such as traffic status detection, traffic sign recognition, vehicular traffic tracking, and so on. Recently, haze removal techniques taken in these particular applications have caught increasing attention in improving the visibility of hazy images in order to make the performances of the intelligent transportation systems more reliable and efficient. However, estimating haze from a single haze image with an actual scene is difficult for visibility restoration methods to accomplish. In order to solve this problem, we propose a haze removal method which requires a combination of two main modules：the haze thickness estimation module and the visibility restoration module. The haze thickness estimation module is based on bi-gamma modification to effectively estimate haze for transmission map. Subsequently, the visibility restoration module utilizes the transmission map to achieve the haze removal. The experimental results demonstrate that the proposed haze removal method can restore the visibility in single haze images more effectively than can other state-of-the-art methods.

並列關鍵字

Dark Channel Prior ； Haze Removal

參考文獻

[1] O. Barinova, V. Lempitsky, and P. Kholi “On Detection of Multiple Object Instances Using Hough Transforms,” IEEE Trans. Pattern Anal. Machine Intell., vol. 34, no. 9, pp. 1773-1784, Sep. 2012.

[2] P. H. Lee, Y. L. Lin, S. C. Chen, C. H. Wu, C. C. Tsai, and Y. P. Hung “Viewpoint-Independent Object Detection Based on Two-Dimensional Contours and Three-Dimensional Sizes,” IEEE Trans. Intelligent Transp. Syst., vol. 12, no. 4, pp. 1599-1608, Dec. 2011.

[3] U. Franke, and S. Heinrich, “Fast Obstacle Detection for Urban Traffic Situations,” IEEE Trans. Intelligent Transp. Syst., vol. 3, no. 3, pp. 173-181, Sep. 2002.

[4] C. Caraffi, S. Cattani, and P. Grisleri, “Off-Road Path and Obstacle Detection Using Decision Networks and Stereo Vision,” IEEE Trans. on Intelligent Transportation Systems., vol. 8, pp. 607–618, Dec. 2007.

[5] S. C. Huang, “An Advanced Motion Detection Algorithm with Video Quality Analysis for Video Surveillance Systems,” IEEE Trans. Circuits Syst. Video Technol., vol. 21, no. 1, pp. 1-14, Jan. 2011.