This paper presents a probabilistic color-based tracking method that improves upon existing deterministic color-based trackers. The proposed method uses a particle filter to handle color clutter and occlusions more effectively. It is based on the principle of comparing the color content of candidate regions to a reference color histogram, but within a probabilistic framework. This approach is more flexible and can be extended in various ways, such as multi-part color modeling, incorporating a background color model, and tracking multiple objects. The method is based on a sequential Monte Carlo framework, which allows for the approximation of the posterior distribution using a set of particles. This enables the tracking of multiple modes of the posterior distribution, which is crucial for handling background distractions and recovering from occlusions. The paper also discusses the extension of the method to multiple objects and the incorporation of a background model, which enhances the robustness of the tracker in scenarios with fixed cameras and known backgrounds. The proposed method is compared to existing deterministic trackers and is shown to be more robust in handling complex tracking scenarios, including those with significant changes in object appearance and shape, and partial or complete occlusions. The method is also demonstrated to be effective in tracking multiple objects, even when they overlap, without swapping identities. The paper also presents an automatic initialization method for tracking faces using skin color models, which is particularly useful in scenarios with still cameras and known backgrounds. The paper concludes that the proposed probabilistic color-based tracking method is a versatile and robust approach that can be extended in various ways to handle different tracking scenarios. It is particularly effective in handling color clutter and occlusions, and is well-suited for applications such as surveillance, video editing, and augmented reality. The method is also shown to be effective in tracking multiple objects, even when they overlap, without identity swapping or loss of track.This paper presents a probabilistic color-based tracking method that improves upon existing deterministic color-based trackers. The proposed method uses a particle filter to handle color clutter and occlusions more effectively. It is based on the principle of comparing the color content of candidate regions to a reference color histogram, but within a probabilistic framework. This approach is more flexible and can be extended in various ways, such as multi-part color modeling, incorporating a background color model, and tracking multiple objects. The method is based on a sequential Monte Carlo framework, which allows for the approximation of the posterior distribution using a set of particles. This enables the tracking of multiple modes of the posterior distribution, which is crucial for handling background distractions and recovering from occlusions. The paper also discusses the extension of the method to multiple objects and the incorporation of a background model, which enhances the robustness of the tracker in scenarios with fixed cameras and known backgrounds. The proposed method is compared to existing deterministic trackers and is shown to be more robust in handling complex tracking scenarios, including those with significant changes in object appearance and shape, and partial or complete occlusions. The method is also demonstrated to be effective in tracking multiple objects, even when they overlap, without swapping identities. The paper also presents an automatic initialization method for tracking faces using skin color models, which is particularly useful in scenarios with still cameras and known backgrounds. The paper concludes that the proposed probabilistic color-based tracking method is a versatile and robust approach that can be extended in various ways to handle different tracking scenarios. It is particularly effective in handling color clutter and occlusions, and is well-suited for applications such as surveillance, video editing, and augmented reality. The method is also shown to be effective in tracking multiple objects, even when they overlap, without identity swapping or loss of track.