Improved background subtraction based on novel thresholding approach

A novel thresholding approach is proposed based on color space model and is embedded into background subtraction. This thresholding approach uses the color distortion and brightness distortion of each pixel to detect the changes in the background. The color distortion is detected by utilizing the position information of color vector in color space. Moreover, this thresholding method is also helpful for removing the moving shadows to some extent. For complex background, a dual-threshold method is proposed in background subtraction to extract object from videos containing complex background. Experimental results on the videos of VSSN 05 and PETS 2006 show the effectiveness of the proposed algorithm.