1. The document proposes a real-time background estimation method for video surveillance using a modified Gaussian Mixture Model (GMM) that estimates the background frame from input video frames rather than classifying pixels as foreground or background.
2. The proposed method models each pixel as a mixture of K Gaussian distributions learned over time and estimates the background, which is then subtracted from the current frame to extract the foreground.
3. Experimental results on real and synthesized videos show the proposed background estimation method performs better foreground extraction in dynamic scenes compared to existing GMM-based methods.