Fast and robust video foreground segmentation for indoor surveillance

This work describes a method of background updating and shadow removal for indoor surveillance. Moving objects can be precisely extracted for various further process procedures such as recognition. Single-Gaussian model which has high computational speed is usually applied to the indoor environment with motionless backgrounds. The pixels of an image are classified as background pixels, moving foreground pixels and motionless foreground pixels, and the Single-Gaussian background model is updated according to the classification of a pixel. The proposed scheme makes the background model respond to environmental changes in time. With the ratio between the foreground pixel value and the background pixel value, pixels are distinguished among foreground, background and shadow. The effectiveness of the proposed method is demonstrated with experiments in an indoor environment.