Description

The main challenge of background subtraction is estimating a robust scene background to handle illumination and static object changes. The background estimation algorithm makes a reference image, which represents the background of the scene view. We assume that the video sequence is captured in an RGB colour space by a stationary video camera. The problem with the RGB colour space is its great sensitivity to sensor noise and lighting condition changes. We first convert the frame from RGB to HSV colour space.

DEMO VIDEO

We propose a novel background extraction algorithm based on an improved mode algorithm to obtain the static background regions. Employing these approaches seeks to obtain a clean static background reference image and apply it to background subtraction. This method eliminates the stationary background, leaving only the desired motion regions. In certain video systems, the acquisition process may be susceptible to erratic changes in illumination, reflection, and noise.

RESULT