In this paper, we describe the development, of a 3D video imaging system for an omnidirectional display in which many people can visualize the same image from different viewpoints at the same time. In order to display such images, in a high quality manner, it is necessary to reduce the discontinuity between adjacent patches during the texture mapping process of reconstructed geometrical shape. To overcome this problem, a view-dependent texture mapping has been used. Such method is only effective for a traditional 2D display. To overcome this limitation, we have developed a view-independent, voxel coloring for an omnidirectional display by using multi-viewpoint video images. In this technique, accurate visibility determination is done in order to diminish the influence of specular reflection, which is the main obstacle for viewpoint independence. We used 12 sets of 300 time-varying frames, each set corresponding to a specific viewpoint, to verify the effectiveness of our approach.
