Bigbackground-based illumination compensation for surveillance video
Journal on Image and Video Processing - Special issue on advanced video-based surveillance
| Hi-index | 0.00 |
Variation in illumination conditions caused by weather,season, time of day, etc., makes the task difficult when buildingsurveillance systems of real world scenes. Especially, castshadows produce troublesome appearance variations to accomplishmonitoring with computer vision techniques, typi-callymoving object tracking from a stationary viewpoint. Torobustly eliminate lighting effects from image sequences as apreprocessing stage for robust video surveillance, we proposea framework based on the idea of intrinsic images. Unlikeprevious methods to derive intrinsic images, we derive time-varyingreflectance images. As a result, we obtain illumina-tionimages that capture only lighting effects on the scene.Using these reflectance images and illumination images, weconstruct an illumination eigenspace to directly estimate theillumination images under the arbitrary lighting conditions ofthe same scene. By canceling out the lighting effects using thisillumination image, robust video surveillance can be accomplished.In this paper, we explain the theory of the frameworkwith simulation results, and apply the framework to real worldmonitoring data set to prove its effectiveness.