Collaborators: Budirijanto Purnomo, Dr. Jonathan Cohen and Dr. Subodh Kumar.
The program takes as input an indexed triangle mesh, and the hardware cache size. The caching scheme uses the FIFO replacement policy. The program outputs a rearranged set of triangles, with (possibly) degenerate triangles. It also outputs a compressed stream of indices, which can be used to reduce the amount of data sent to the graphics hardware per frame. The code is written in C/C++, and uses OpenGL and glvu API for rendering the model.
Collaborators: Budirijanto Purnomo, Dr. Shankar Krishnan, Dr. Jonathan Cohen and Dr. Subodh Kumar.
The system has two main phases. The first phase, a pre-computation phase, takes as input a triangular model, and computes connected components, and various levels of detail of the components formed. The viewing area is then divided into cells, and the cell-based visibility is computed, and the obtained data compressed, encoded and stored. The second phase, a run-time phases, computes the cell in which the user lies, and displays the visible part at the correct level of detail. We have run the system on various models including the UNC Powerplant model (12 million triangles.) We are currently trying to render the Boeing model (350 million triangles.) The code is written in C/C++, and uses OpenGL and glvu API for rendering the model.
Collaborators: Budirijanto Purnomo, Dr. Shankar Krishnan, Dr.Jonathan Cohen and Dr. Subodh Kumar.
Given an input triangular mesh, and a set of potential occluders, and a view-cell, the program outputs the set of triangles (conservatively), which are visible from any point in the view-cell. This system in used as a part of the vLOD system described above. We use the Nvidia Occlusion query to speed up the visibility computation. The code is written in C/C++, and uses OpenGL and glvu API for rendering the model.
Collaborators: Sudhir Vishwanath, Dr. Jonathan Cohen and Dr. Subodh Kumar.
The system has two main phases. The first phase, a pre-computation phase, takes as input a voxelised grid, and computes the set of iso-values at which the iso-surface can potentially undergo a change in topology. The second phase, the run-time phase, determines the cells intersecting the isosurface, and displays the isosurface. The code is written in C/C++, and uses OpenGL and glvu API for rendering the model.
Collaborator: Dr. Subodh Kumar.
The system takes an input a model comprising of parametric surfaces. The pre-computation phase triangulates (delaunay) each patch, and computes and stores a set of points with each patch. During the run-time phase, a set of the points is selected, and displayed using the OpenGL glPoint API, with a certain point size. The number of points are chosen to satisfy a pre-defined user threshold to ensure interactivity. The code is written in C/C++, and uses OpenGL and glvu API for rendering the model.
Collaborators: Jason Corso and Dr. Allison Okamura.
The system takes a set of points and fits a parametric (rational Bezier) surface. The various physical properties (like spring constants, damping factors, etc) are embedded in the higher dimensions of the surface. The user interacts with the surface using a 6-DOF Phantom haptic device. We perform collision detection, and force computation at a frequency of 1Khz, updating the image in real-time (at 15-20 Hz.) The code is written in Visual C++ and uses the OpenGL API for rendering the model.
Collaborator: Dr. Subodh Kumar.
The system takes an input a model comprising of parametric surfaces (with possibly trimming curves.) The pre-computation phase triangulates (delaunay) each patch, and computes and stores a set of points with each patch. During the run-time phase, a set of the points is selected, and incrementally triangulated, so that the displayed mesh is within a user-specified screen-space error threshold. In case the pre-computed set of points is exhausted, the regions of a patch having high deviation are uniformly tessellated. The code is written in C/C++, and uses OpenGL and glvu API for rendering the model.
Collaborators: Harpal Singh Bassali, Saurabh Agarwal, Dr. Pradeep Dubey and Dr. Alok Aggarwal.
The system takes an input a 2D image, and embeds a compression tolerant invisible watermark in the image. The watermark survives lossy compression techniques (like JPEG.) The code is written is C/C++.