We propose a two-stage real-time tracking algorithm for an active camera system having pan-tilt-zoom functions. The algorithm is based on the assumption that a human head has an elliptical shape and that its model color histogram has been acquired in advance. The algorithm consists of two stages, a color-based convergence stage for fast and reliable tracking and a refinement stage for accurate tracking based on multimodal information. In the first color convergence stage, we roughly estimate the target position by using the mean-shift method based on the histogram similarity between the model and a candidate ellipse. To better predict the initial position for the mean shift, the global motion is compensated; to enhance reliability of the mean shift, the model histogram is appropriately updated by referring to the target histogram in the previous frame. In the subsequent refinement stage, we refine the position and size of the ellipse obtained at the first stage by using multimodal information such as color, shape, and quasi-spatial information. In particular, to quantify the quasi-spatial information, we use a spatial color histogram obtained by properly dividing the ellipse into two regions. Extensive experiments verify that the proposed algorithm robustly tracks the head, even when the subject moves quickly, the head size changes drastically, or the background has many clusters and/or distracting colors. Also, the proposed algorithm can perform real-time tracking with a processing speed of about 10 fps on a standard PC.
In this paper, we propose a robust real-time head tracking algorithm using a pan-tilt-zoom camera. In the algorithm, the shape of the head is assumed as an ellipse and a model color histogram is acquired in advance. Then, in the first frame, a user defines the position and scale of a head. In the following frame, we consider the ellipse selected in the previous frame as the initial position, and apply the mean-shift procedure to move the position to the target center where the color histogram similarity to the reference one is maximized. Here, the reference color histogram is adaptively updated from the model color histogram by using the one in the previous frame. Then, by using gradient information, the position and scale of the ellipse are further refined. Large background motion often makes the initial position not converge to the target position. To alleviate this problem, we estimate a reliable initial position by compensating the background motion. Here, to reduce the computational burden of motion estimation, we use the vertical and horizontal 1-D projection dataset. Extensive experiments show that a head is well tracked even when a person moves fast and the scale of the head changes drastically.