Title :
Intelligent video defogging technology based on covariance and perceptron
Author :
Longli Li ; Qing Liu ; Jianming Guo ; Yanfan Xiong
Author_Institution :
Sch. of Autom., Wuhan Univ. of Technol., Wuhan, China
Abstract :
As the algorithm of video defogging must meet real-time requirement, a new method is proposed, it uses the covariance matrix of multi-feature combination to describe image features, and combines with perceptron model to intelligently detect foggy scenes. Because the backgrounds of industrial video images generally change slowly, Gaussian mixture modeling is used to get foregrounds. The transmission of dark channel prior is updated according to the foreground. Then each frame is restored directly according to the newer transmission. The defogging algorithm greatly reduces the running time. It achieves the purpose of video defogging. Experimental results show that the algorithm has a high accuracy on detecting foggy scenes. The algorithm of video defogging proposed can meet the industrial real-time requirements and ensure spatial and temporal consistency of video.
Keywords :
Gaussian processes; covariance matrices; object detection; perceptrons; video signal processing; Gaussian mixture modeling; covariance matrix; dark channel; foggy scene detection; image features; industrial video images; intelligent video defogging technology; multifeature combination; perceptron model; spatial consistency; temporal consistency; Classification algorithms; Computational modeling; Covariance matrix; Image color analysis; Real time systems; Streaming media; Support vector machine classification; covariance; dark channel prior; defogging; multi-feature combination; perceptron; video;
Conference_Titel :
Natural Computation (ICNC), 2011 Seventh International Conference on
Conference_Location :
Shanghai
Print_ISBN :
978-1-4244-9950-2
DOI :
10.1109/ICNC.2011.6021916