Compression of the Concentric Mosaic

Figure 1. Capturing device of the concentric mosaic.

This project is for the compressing and browsing of virtual environment.

The concentric mosaic is an easy way to represent a realistic environment. The capturing facility of the concentric mosaic can be shown in Figure 1. We just rotate a single camera at the end of a level beam, with the camera pointing outward and shooting images as the beam rotates. The resultant dataset is a video sequence. For a typical realistic environment with large depth variation, 900 to 1500 shots have to be captured in a circle to render the scene properly without alias. This leads to a dataset of several hundreds of megabytes.

Figure 2. Rendering of concentric mosaic

Rendering the concentric mosaic involves reassembling slits from existing photo shots. No 3D or depth information of the scene is needed to reconstruct a novel virtual view of the environment. Shown in Figure 2, let P be a novel viewpoint and AB be the field of view to be rendered. We split the view into multiple vertical slits, and render each slit independently. Let the slit PV be a rendered slit. We simply search for the slit P'V in the captured dataset and use the content of P'V to replace PV, where P' is the intersection between ray PV and the camera path. The basic hypothesis here is that the intensity of any ray does not change along a straight line unless blocked. Therefore, what is rendered at PV can be recovered from that is observed in P'V.

For application of the concentric mosaic, the large dataset has to be efficiently compressed. Moreover, since only a portion of the data is used to render each concentric mosaic view, it is imperative that the compressed data to be formed in an indexable fashion. That is, portion of the compressed concentric mosaic should be accessed and delivered in a timely fashion.

Existing researches on concentric mosaic can be classified into 3 categories, block coder, reference frame coder and 3D wavelet coder. I have written a survey paper on the subject, and it can be found below. I have also investigated efficient reference frame coder and 3D wavelet coder. If you are interested in our work, please follow the link.

• J. Li, H. Shum and Y. Zhang, "On the compression of the image based rendering scene: a comparison among block, reference and wavelet coders", International Journal on Image and Graphics, Vol. 1, No. 1, pp.45-61, 2001.

• J. Li, H. Shum and Y. Zhang, "On the compression of image based rendering scene", IEEE International Conference on Image Processing (ICIP 2000), #3307, Vancouver, BC, Sept. 2000.
• Compression of concentric mosaic by reference frame approaches
• Compression of concentric mosaic by 3D wavelet approaches

©Copyright by Jin Li, June 22, 2001