A new particle filter, Kernel Particle Filter (KPF), is proposed for visual tracking for multiple objects in image sequences. The KPF invokes kernels to form a continuous estimate of the posterior density function and allocates particles based on the gradient derived from the kernel density estimate. A data association technique is also proposed to resolve the motion correspondence ambiguities that arise when multiple objects are present. The data association technique introduces minimal amount of computation by making use of the intermediate results obtained in particle allocation. We show that KPF performs robust multiple object tracking with improved sampling efficiency.