Parallelisms in MPEG and its applications to 3-D visualization

Techniques for visualization of 3-D objects must present to the user photo-realistic images in addition to offering various 3-D manipulations including rotation and translation as well as panning and zooming capabilities. The image-based visualization can satisfy these requirements at low-cost by capturing images of the 3-D object at all view angles. This technique naturally offers photo-realistic 3-D visualization, since the viewing software-while the user manipulates the 3-D object in real-time-simply displays the corresponding frame. However, due to the large number of captured images, typically 300 to 600 cuts, the data must be efficiently compressed; and at the same time the compressed data must allow real-time decoding to give the effect of 3-D manipulation to the viewer. As neighboring frames (in close proximity in terms of the view angle) will almost always be similar, we have opted to use MPEG to compress the captured data. The MPEG standard offers a multitude parallelism that can be exploited and in this paper, we discuss various parallelisms in MPEG applicable to our image-based 3-D visualization for achieving photo-realistic visualization and manipulation of 3-D objects.