Tree Morphing
The prime focus of this work is morphing a source tree into a target tree after computing the proper matching between the nodes of the two trees. The number of nodes/edges of the tree may not be same.
Applications : animation/ tree visualization.
Analysis of Self-overlapping curves
Self-overlapping curves are obtained by stretching and overlapping a disk without twisting it. The boundary of such a morphed disk is called a self-overlapping curve. The typical questions that can be asked, given such a curve, are:
1. Can it be segmented into smaller disks bounded by simple polygons?
2. Does this curve bound a single disk or multiple, characteristically different disks?
3. Can such a disk be embedded in 3D such that the underlying surface does not intersect itself?
4. Given any closed loop curve can it be determined in polynomial time whether it bounds a disk?
This work focusses on the above questions and an in depth analysis of the class of curves.
Application : animation, compression of graphical models,
3D reconstruction.
paper
Morphing self-overlapping curves
Given two self-overlapping polygons, they are smoothly transformed from one to another. This is the first algorithm to achieve this type of morphing. Typical commercial animation packages like Adobe Flash cannot handle curves beyond those that are bounded by simple polygons. This work specifically addresses this problem and solves it for a class of self-overlapping polygons.
Applications: Morphing, shape tweening
paper
Interactive Volume Rendering
As medical imaging techniques are becoming more and more sophisticated, the need to render huge CT scanned models at interactive frame rates are becoming more and more important. With this goal in mind, an interactive volume renderer is designed that is capable of rendering huge tensor models at interactive frame rates, leveraging modern GPU capabilities (designed using CUDA). The renderer allows the user to zoom in/out, pan, rotate, and interactively change the transfer function.
The present focus of this work is to compress huge volume data, send this data to GPU and then perform decompression on the GPU at run-time.
Interactive Polygonal Model Rendering
( done as an intern at MSC Software, summer 2012)
This work focusses on huge polygonal model rendering (~ several 100 million points/triangles) at interactive frame rates using GPU compuations. The model is divided into several smaller blocks and depending upon the current viewpoint the level of detail (LOD) of each block is computed before rendering. Based on the LOD, the connectivity of the points have to be recomputed on a per frame basis.
Segmentation and Tracking of Neuro Stem Cells
This work, done in collaboration with the dept. of Neuro Science, UCI, focusses on the segmentation and tracking of human stem cells over a series of time frames. Segmentation is performed using a variant of the Watershed algorithm, and geometric algorithms like constrained Delaunay Triangulation are used to detect important events like mitosis, cell death etc.
paper