Estimating adaptive coefficients of evolving GMMs for online video segmentation

A new, online, evolving video segmentation algorithm is presented in this paper. The proposed method segments each video frame using an evolving Gaussian mixture model (GMM) whose adaptive coefficient is automatically adjusted to cater for abrupt changes between consecutive frames. The proposed method is tested against another algorithm, which keeps the adaptive coefficient constant. The comparison shows the advantage of altering the value of the adaptive coefficient according to change in the scene.