Publications

Abstract : Rendering of natural scenes has been widely discussed by many researchers for a long time due to its numerous applications. But still the main challenge is complexity in geometry and memory unavailability in current hardware platforms. Natural scenes from real world contain a huge number of small details that are hard to model, take a lot of time to render and require a huge amount of memory. We address this problem by following the principle of self similarity or fractal geometry in the natural scenes. For evaluating the feasibility of fractal based image rendering in different dimensions, we have first considered the 2D structure, Mandelbrot set that has gained wide recognition both in mathematical and graphical domains because of its appeal and complex structure. In our work, we have examined the serial algorithm of this set and devised a parallel algorithm for the implementation on a massive parallel graphics processing unit (GPU) using the computer unified device architecture (CUDA) programming model. We have also extended our approach from 2D structure like Mandelbrot set to 3D real world example of terrain rendering. Performance is evaluated in terms of execution time and observed that a parallel implementation of the method on a GeForce GTS 450 GPU is on an average 2X times faster than its sequential implementation.