This paper presents a computer vision algorithm called CAMSHIFT for face tracking, which is used in a perceptual user interface. The algorithm is based on the mean shift algorithm, which is a non-parametric technique for finding the mode of probability distributions. The CAMSHIFT algorithm is modified to adapt dynamically to changing color probability distributions derived from video frame sequences. It is used to track human faces in real time and to control commercial computer games and explore 3D graphic worlds. The paper discusses the development of a 4-degree of freedom color object tracker and its application to face tracking. It describes the use of color histograms to represent color probability distributions and the use of the HSV color system to separate hue, saturation, and brightness. The algorithm is robust to noise, distractors, and lighting variations, and it can track faces in the presence of other faces, hand movements, and occlusions. The paper also discusses the performance of CAMSHIFT in terms of computational efficiency and its ability to handle noise and distractions. It compares the accuracy of CAMSHIFT with a Polhemus tracker and shows that CAMSHIFT performs well in tracking faces in noisy environments. The algorithm is also shown to be robust to transient occlusions and can track faces even when parts of the face are obscured. CAMSHIFT is used as a perceptual user interface for controlling computer games and exploring 3D graphic worlds. It is a simple and computationally efficient algorithm that tracks four degrees of freedom (X, Y, Z, and head roll). The algorithm is designed to be part of a larger tracking system that allows computers to track and understand human motion, pose, and tool use. It is also designed to be used in the future with more advanced tracking systems that provide more robust tracking, posture understanding, gesture and face recognition, and object understanding.This paper presents a computer vision algorithm called CAMSHIFT for face tracking, which is used in a perceptual user interface. The algorithm is based on the mean shift algorithm, which is a non-parametric technique for finding the mode of probability distributions. The CAMSHIFT algorithm is modified to adapt dynamically to changing color probability distributions derived from video frame sequences. It is used to track human faces in real time and to control commercial computer games and explore 3D graphic worlds. The paper discusses the development of a 4-degree of freedom color object tracker and its application to face tracking. It describes the use of color histograms to represent color probability distributions and the use of the HSV color system to separate hue, saturation, and brightness. The algorithm is robust to noise, distractors, and lighting variations, and it can track faces in the presence of other faces, hand movements, and occlusions. The paper also discusses the performance of CAMSHIFT in terms of computational efficiency and its ability to handle noise and distractions. It compares the accuracy of CAMSHIFT with a Polhemus tracker and shows that CAMSHIFT performs well in tracking faces in noisy environments. The algorithm is also shown to be robust to transient occlusions and can track faces even when parts of the face are obscured. CAMSHIFT is used as a perceptual user interface for controlling computer games and exploring 3D graphic worlds. It is a simple and computationally efficient algorithm that tracks four degrees of freedom (X, Y, Z, and head roll). The algorithm is designed to be part of a larger tracking system that allows computers to track and understand human motion, pose, and tool use. It is also designed to be used in the future with more advanced tracking systems that provide more robust tracking, posture understanding, gesture and face recognition, and object understanding.