A new approach is proposed for matching uniform and nonuniform regions of moving objects extracted from an image sequence. The nonstationary regions in the difference picture, the absolute difference of two images, are first extracted. Their shapes are superimposed on the segmented regions of the original images. According to the ratio of areas inside and outside the intersected nonstationary regions, the segmented regions are classified as dynamic or static. The dynamic regions clustered together are assumed to be of the same moving object and are called object regions. An inter-object region matching procedure is applied to the object regions in two consecutive images. This procedure takes gray level, area and moment as the features of the regions to define the cost function. Finally, an intra-object region matching procedure based on the relaxation method is performed over clustered regions. Some experimental results demonstrate the efficiency of this approach.