摄像头表面脏污引起的图像质量下降，会造成图像识别算法精度降低，最终对智能汽车驾驶安全造成很大的影响。汽车在实际道路行驶过程中拍摄到的图像都是无参考图像，本文将无参考图像质量评价方法中的相关系数、边缘信息、频域能量和平均信息熵等方法引入到道路图像的识别中，对在道路上获取的清晰序列和脏污序列图像进行判别，结果表明，该方法对清晰图像和脏污图像具有一定的区分能力。
The degradation of image quality caused by dirty in lens surface will reduce the accuracy of image recognition algorithm, which eventually has a great impact on the driving safety of intelligent driving assistance system (IDAS). Images captured on the road are no reference images. In this article, methods of no reference image quality assessment such as correlation coefficient, edge information, frequency domain energy and average information entropy will be used in the image recognition. Those methods are used to distinguish the clear images and dirty images obtained on the road. The results show that those methods have a certain degree of distinction between clear images and dirty images.
Gu, J., Belhumeur, P., Nayar, S., et al. (2009) Removing Image Artifacts Due to Dirty Camera Lenses and Thin Occluders. ACM Transactions on Graphics, 28, 1-10. https://doi.org/10.1145/1618452.1618490
[4]
Fujii, Y., Ohta, N., Ito, T., et al. (2006) Image Restoration for Security Cameras with Dirty Lens under Oblique Illumination. IEEE International Workshop on Imagining Systems & Techniques.
[5]
Einecke, N., Gandhi, H. and Jorg, D. (2014) Detection of Camera Artifacts from Camera Images. IEEE International Conference on Intelligent Transportation Systems, Qingdao, 8-11 October 2014, 603-610.
https://doi.org/10.1109/ITSC.2014.6957756
