Naveen Kumar Bolla

life, work and photography

Main menu: home | work | projects | resume | photography | misc | contact

Keywords: Algebraic Splats, Point Based Graphics, Implicit Surfaces, Ray Tracing, GPU Computing, Depthmap Compression [other projects ]

Algebraic Splats Representation for Point Based Models

In point-based representations a large number of surfels (linear primitives) are needed to convey the exact shape. Higher-order approximations of the local neighbourhood have the potential to represent the shape using fewer primitives, simultaneously achieving higher rendering speeds. In this project, we proposed algebraic splats as a basic primitive of representation for point based models. Quadratic and cubic splats provide good quality and high rendering speed using far fewer primitives on a wide range of models. They can also be rendered fast using ray tracing on modern GPUs. Our representation reduces the number of primitives needed by a factor 20 to 30 on most models and by a factor of over 100 on dense models like David with little or no drop in visual quality. We are able to render models like David at upwards of 100 frames per second on a commodity GPU using algebraic splats [PDF]

Algebraic Splats for High-Quality Rendering of Points

Algebraic splats (Quadratic and Cubic) as a basic primitive of representation for point based models have the potential to represent the shape using fewer primitives, simultaneously achieving higher rendering speeds with better shading. Algebraic splats can be made more effective with culling and Level-of-detail scheme. Our representation supports octree-based representation for continuous level of detail. We present a real-time ray tracing algorithm with culling, adaptive anti-aliasing and shadows for large point sets at consistent speed. We employed a two-pass GPU algorithm that ray-traces the algebraic splats and blends them using a Gaussian weighting scheme for smooth appearance.

Ray Casting Implicit Surfaces using CUDA

A ray-tracing procedure to render general implicit surfaces efficiently on the GPU is presented here. This adaptive marching points algorithm fits the SIMD architecture of the GPU results in high performance.Implicit surfaces are ray-traced using CUDA on GPUs. Interesting thing about this algorithm is that it involves a lot of number crunching and a very few data movements operations. This offered us an ideal application for CUDA computation model. The amount of shared memory required is zero. We are able to extract around 700 GFLOPS of performance from Nvidia GTX280.

Proxy Based Compression of Multiple Depth Images

Compression of multiple depth-maps of a scene has not been worked on a lot. Depth-maps differ from images qualitatively and new methods are needed to compress them. While images can be compressed using lossy methods like JPEG, we can't afford to compress Depth Images because a lossy method would effect the true depth of the scene and thus the whole structure of the scene would get deformed. Hence, we need a novel method to compress depth efficiently keeping the depth values viable. We used geometric proxy of the scene to compress multiple depth maps of the scene.The residue depth maps are compressed using various compression algorithms. A lot of experiments are conducted on the wide variety of the models with proxies of various LOD’s, various camera configurations, various encoding and sampling of the depth maps. [PDF]