Enhancement of robust foreground detection through masked greyworld and color co-occurrence approach

Total robustness to illumination change is difficult to solve just using a single approach. Our approach to improvement in accuracy and robustness to illumination change is by fusing two different methods, namely, color constancy and color co-occurrence. A “grey world” assumption is used to handle sudden illumination change while color cooccurrence modelling is used to handle gradual illumination change. The novelty of our approach is the introduction of a mask and a probabilistic approach for adjustment of the grey constant values. In addition, because of our novel color cooccurrence approach, our algorithm can distinguish between “moving” background and moving foreground. The results prove that our method perform the best among the compared methods and manage to reduce the error rate by half compared to the Renno et al. method. In conclusion, the fusion of the approach manages to improve robustness to illumination change and moving background.