Compression of Concentric Mosaic with Rebinning of Slits (ROSS)

Concentric mosaic offers a quick solution for the construction and navigation of a virtual environment. However, the associ-ated huge amount of data presents a heavy burden for its application. In this paper, we propose the rebinning of slits (ROSS) approach for the compression of concentric mosaic. The proposed scheme decorrelates the 3D concentric mosaic data set by rebinning the vertical slits from all captured image shots into a 2D panorama, and then compresses the panorama with a highly efficient still image coder, such as JPEG2000. Unlike typical video coders that use motion compensation to decorrelate multiple frames, the rebinning converts 3D spatial correlation into stronger 2D correlation. Experimental results show that ROSS outperforms MPEG-2 coding of concentric mosaic by an average of 1.0dB. mosaic compression algorithm. To facilitate JIT rendering, the bitstream needs to be random accessible and the decoder should be reasonably fast to accommodate the real-time decoding need. In [1], a spatial vector quantization (SVQ) scheme has been proposed to compress concentric mosaic. SVQ is fast in decoding, and the compressed SVQ index can be easily accessed at arbitrary location. However, SVQ is complex in the encoding stage, and the compression performance of SVQ is limited, e.g., 12:1. If we raise the compression ratio higher, the scene quality degrades quickly. An alternative approach is to treat the concentric mosaic as a video and compress it with a video