Image-Space Dynamic Transparency for Improved Object Discovery in 3D Environments

We present an image-space algorithm for dynamic transparency with the purpose of supporting object discovery and access in information-rich 3D visualization environments. The algorithm is based on multiple rendering passes and detects instances of object occlusion in the image-space using the fragment shader capabilities of modern programmable graphics hardware, creating alpha maps of opacity gradients around the occluded objects. In essence, the effect is somewhat similar to X-ray vision of a superhero. We have implemented a prototype version of our algorithm with real-time rendering performance using a number of optimizations and speedups on current graphics hardware. To evaluate its use, we have also implemented three different example applications portraying different scenarios, including abstract visualization, virtual walkthrough, and gaming. Preliminary results from an empirical user study comparing our technique to standard viewpoint controls indicate that our technique is superior in regards to object discovery efficiency. These results hold over both completion times as well as correctness of a visual search task