[katieHaylor (OP)]

Phillip says:

Within the last few decades, someone discovered that the quality of Computer Generated Graphics is immensely improved and time required much reduced by using fractals - at least I believe so.

Is that true and if so, why?

What do you think?

[Telgin]

I'm not aware of any improvement brought about specifically by fractals, but I can add some commentary depending on which domain of computer graphics you're asking about in particular.  I'll assume you mean photorealistic graphics like used in movies, as opposed to the kinds of graphics used in computer games.

Actually, either way, there are generally two things that lead to better (and faster) computer graphics:

More Processing Power

Pretty obvious, but more processing power lets us do more.  I believe it was John Carmack, one of the main programmers for id software back in the day and inventor of a lot of game graphics techniques, once said that we can do lots of optimizations and tricks, but at the end of the day it's raw power that opens up the frontier and lets us do new things.

For photorealistic rendering, more power lets you run more complicated shading algorithms (see below) and also lets you perform higher resolution simulations of lighting.  Traditionally, photorealistic rendering is done by ray tracing, which literally involves simulating the path of "photons" emitted from light sources as they bounce around and illuminate a scene.  More processing power lets us simulate more of these photons, lets us track them over more bounces, and so on.  That increases both the quality of the lighting and the speed of the rendering.

Games generally do not use raytracing for lighting and do a lot of other complicated hacks to do things quickly, but even still, more processing power lets us do more complicated things that look better.

Algorithms

This is where fractals would come in, if anywhere, but again I'm not aware of any use of fractals that has  revolutionized anything.

For photorealistic rendering, understanding how materials react to light is the key to making them look accurate.  Renderers have to understand how light is absorbed, reflected, bent and so on by a material to produce its final color on screen.

One relatively recent innovation here is the notion of subsurface scattering, which is a way that some materials like skin behave under light.  I don't know a lot about its specifics, but it models the semi translucent quality of skin (such that blue from veins shows up, for example), and it dramatically improved the quality of rendered skin.  The lack of subsurface scattering is what leads to rendered skin having a plastic or rubber look.

After Thoughts on Fractals

There actually may be a specific benefit to using fractals, now that I think about it, but it really only relates to the speed of rendering and mostly matters for computer games.  Without getting into very nitty gritty details, if you can express a noise or fractal as a simple mathematical function, you can save memory bandwidth on a graphics card by computing that fractal per pixel instead of looking it up from a texture.  Some games use noise maps like this to make surfaces look more interesting without having to manually paint textures.

This isn't particularly new though, and the computation can outweigh the performance gains of just encoding it in a texture, especially if the fractal function requires input from neighboring cells or pixels.

[Monicaiz]

The main forum of the forum is the channel to tell it.