Bayesian Illumination Invariant Motion Detection Using a New Adaptive Threshold

Detection of moving objects plays a very important role in video processing of many vision applications. The vision systems constituting of image processing algorithms are widely implemented and exploited in various applications such as traffic control, video surveillance of objects, and etc. Most basic change detection algorithms assume that the illumination of relative scene will remain constant. However, this assumption is not necessarily valid particularly outside a well-controlled laboratory setup. In this paper, a novel illumination-invariant method of change detection has been presented which is insensitive to varying illumination. The proposed algorithm has been based on a Bayesian change detection method using a threshold that is adaptively controlled so that the effects of the illumination of image frames are minimized. Proposed adaptation algorithm employs distance measure.

Keywords:

Motion , adaptive threshold , Bayesian , distance measure , illumination