Gallery

abstract

My interests within the computer graphics field have always been with the rendering side of the pipeline. Since the second year of my degree, I have been forming a solid basis of knowledge regarding the use of shaders and have accumulated a reasonable amount of programming experience. During the group project, in the second year, my colleagues and I struggled to get all of our renders through the mental ray[12] for maya[13] implementation for maya 5. Masses of geometry caused the rendering to be extremely slow and with a lot of objects moving repeatedly in and out of shot, a great reduction in the polygon count proved difficult. It was not feasible to make the objects completely disappear when they were not visible because mental ray[12] requires all the objects to be present, especially when using global illumination and final gathering. This situation was the ‘spark’ that initiated my idea for the innovations project.

During the summer, I spent several months gaining experience while I completed a placement in the research and development team of The Moving Picture Company, situated in London. Rendering takes a great deal of the company’s time, as modern day computer architectures (x86), have got a limited amount of memory address space. (4GB) If any scene required more memory, the render would inevitably fail. It was the task of the research and development department to find clever ways of reducing the memory needed to render, without unduly compromising image quality.

If one is aware of the recent developments in the computer hardware world, one might be know that the 4GB threshold is soon to disappear, with the widespread development of x86-64 bit machines. In the near future, it will be possible to access over 4GB of data on standard machines. However, with the prices of RAM being as they are, it will still be very expensive for companies to equip every computer in a renderfarm with such amounts of RAM.

Although the architecture used for rendering within the industry differs noticeably from that of a six man group project, the problem remains the same. How to render more efficiently. Pixar’s prman[11] is in essence a scanline renderer and as a result it can handle much more geometry in a scene than a ray trace renderer like mental ray. It would therefore be a very innovative concept to develop a tool for use within maya, that would allow mental ray to handle substantially increased amounts of geometry without compromising render times and image quality.

To realize this idea, I started looking at polygon reduction. Polygon reduction is a well-researched topic, due to the obvious benefits of having a well optimized mesh for rendering or manipulating. However, I had not come across a tool that would reduce the object, in accordance with user settings, and the render camera position. When one takes the camera position into account, certain decisions can be made about the mesh, which do not produce the best reduction mathematically but do produce the best visual result.

SVanDerSteen.zip

SVanDerSteen.zip