Background Subtraction: Creating an ensemble structure of mixture of gaussians

Question

I want to improve the background subtraction in a video by modifying the standard MOG algorithm and creating an ensemble structure of Gaussians modeled on various features of a frame like normalized chromaticity values, spatial gradient and texture. I hope to reduce the illumination effects by including the spatial features. What would be a good way to implement the same preferably using openCV? The ensemble structure would either be based on AdaBoost or a simple weighted combination of all Gaussian models.