Video background replacement using a genetic algorithm
Yangmi Lim, Jinwan Park · Optical Engineering · 2008
Recent statistical methods of video background replacement are robust enough to operate in dynamic environments, but generally require very large computational resources and still have difficulty in clear segmentation of objects. We use a simpler running-average method to model a changing background, and a single global threshold vector, optimized by a genetic algorithm, instead of pixel-by-pixel thresholds. A fitness function is trained to evaluate segmentations by penalizing incorrectly recognized regions. Experimental results on real images show that our new approach outperforms an existing method based on a mixture of Gaussians.