A Fast Convergent Adaptive-K Mixture-Of-Gaussian Model for Video Object Segmentation

Abstract

It is a key step to extract moving objects from background for computer vision applications. GMM based methods are the most commonly used technique for background subtraction in video sequence. However, how to establish efficient and precision background model with fast convergence rate is a Research-Worthy problem. In this paper, an effective scheme is proposed to accelerate the convergence rate of Adaptive-K Gaussian Mixture Model (AKGMM). The AKGMM algorithm alters the dimension of the parameter space at each pixel based on the changing frequency of pixel value. The number of GMM reflects the complexity of pattern at the pixel. An improved learning method is proposed for Gaussian Mixture Model. An adaptive learning rate is calculated for each Gaussian at every frame for speeding up the convergence without compromising model stability. Experimental results demonstrated that the proposed method gets a faster convergence while maintaining good robustness against complex environment compared to a conventional method.