Issue No. 01 - January-February (2004 vol. 10)
<p><b>Abstract</b>-We propose a novel 2D representation for 3D visibility sorting, the <it>Binary-Space-Partitioned Image (BSPI)</it>, to accelerate real-time image-based rendering. BSPI is an efficient 2D realization of a 3D BSP tree, which is commonly used in computer graphics for time-critical visibility sorting. Since the overall structure of a BSP tree is encoded in a BSPI, traversing a BSPI is comparable to traversing the corresponding BSP tree. BSPI performs visibility sorting efficiently and accurately in the 2D image space by warping the reference image triangle-by-triangle instead of pixel-by-pixel. Multiple BSPIs can be combined to solve “disocclusion,” when an occluded portion of the scene becomes visible at a novel viewpoint. Our method is highly automatic, including a tensor voting preprocessing step that generates candidate image partition lines for BSPIs, filters the noisy input data by rejecting outliers, and interpolates missing information. Our system has been applied to a variety of real data, including stereo, motion, and range images.</p>
Visibility sorting, binary space partitioning, image-based rendering, segmentation.
Wai-Shun Tong, Chi-Keung Tang, Andrew J. Hanson, Tien-Tsin Wong, Chi-Wing Fu, "Binary-Space-Partitioned Images for Resolving Image-Based Visibility", IEEE Transactions on Visualization & Computer Graphics, vol. 10, no. , pp. 58-71, January-February 2004, doi:10.1109/TVCG.2004.1260758