Here we propose the method for tracking of the selected object in video sequence in the absence of a-prior information about background evolution, as well as size and a form of an object. Under such conditions a contouring of an object has to be performed prior to a tracking. Maximum likelihood method is employed for efficient contouring. Estimation of likelihood function, based on residual functions of all parts of frame, is proposed. We implemented an iterative algorithm for maximizing the estimated likelihood function. Developed algorithm showed high performance and stability in a wide spectrum of videos.