OCR Text |
Show RESEARCH POSTERS ON THE HILL SPRING 2007 105 Procedural Generation: Computers Generating Art Eli Ribble and Pete Shirley School of Computing The top two images of this poster offer a very interesting contrast. On the left is a fractal image, composed of a simple mathematical expression and some simple color rules. It has interesting, highly detailed structure, and if we were viewing it on a computer screen with an appropriate program, we would see that we can zoom in on the image indefinitely. It has an infinitely high resolution. On the right we have 3 simple circles. Given information on exactly what colors to use, exactly how big to make the circles, and where their center points are, any person could recreate this image. The computer, on the other hand, has no notion of circles, and therefore couldn't recreate this image as a human would, using a handful of numbers and some colors. In order for the computer to store this image and display it on screen, the computer has to know every dot of color. That's a list of 786,432 color triplets, one triplet for every dot in this 1024x768 image. That means that while the fractal image on the left is infinitely more complex and interesting, the image on the right takes up many more times the amount of storage space on a computer. On top of that, if you were to ask the computer to enlarge the image, eventually you would no longer have circles, but rather blobs of color composed of tiny squares. That is, after all, how the computer creates the image - out of 786,432 tiny squares. The question then becomes: can we program a computer to understand the idea of circles? The answer is yes. Using a technique called procedural generation, we can cause the computer to generate images using mathematical procedures, rather than a list of dots. A fractal is a simple and well-known procedurally generated image. Fractals have been known to mathematicians for many years, but not until the recent advent of computers have they been able to actually create images from fractal equations. Of course, while fractals are interesting to look at, they do not have as much practical use to us unless we can do something with them to make images that mimic real life. The next image shows a wood grain texture. These are used often in computer graphics, and normally the wood grain is just a picture of an actual piece of wood. The portion of the image on the left is just that- a picture of real wood. On the right is a wood texture that was created procedurally by the computer. The process is broken down on the next image. Different simple bases are transformed using simple mathematical processes, then combined together into an image that is surprisingly similar to the original. Because this texture is created using mathematical algorithms rather than a list of color points, it has some very useful properties. First, like the fractal image, it has infinite resolution and can be enlarged indefinitely. Second, it is tileable, or in other words it can be placed next to itself with no visible seams. Finally, the data used to create the image is extremely compact. In fact, the wood grained texture can be stored in less space on a computer than this entire paragraph. Wood textures aren't the only thing that can be procedurally generated, however. Anything that is represented in a computer can be generated by the computer rather than stored. The next image shows a small statue of a Buddha. This image was created by taking an actual statue, scanning it with a laser, creating a list of thousands of triangles from the datapoints that approximate the statue, and then running those triangles through a 3d image generator. Because the statue is represented in the computer by thousands of triangles, it again takes a lot of space and has a limited resolution. The final image of the waterfall, on the other hand, is procedurally generated. In fact, it is a snapshot of a 4 minute video that is complete with titlescreen, credits and soundtrack, all generated by the computer. The entire video fits in 64K, or 65,024 bytes. That means that the still image of the video, stored the traditional way, is about 8 times as large as the entire video. This video was produced by one of the many hobbyist groups in Europe that makes up the demoscene. The demoscene has existed for two decades as a group who competes in using procedural generation and computer expertise to create stunning works of computer art. Procedural generation technology in Europe is many years in advance of the United States because of these competitions, and only in the past few years has the United States begun to discover the depth of the work that has been done there. The SCI Institute at the University of Utah is only now beginning to research how it can marry its cutting-edge real-time ray-tracing image generating software with procedural generation techniques to create interactive images for scientific computing This research is supported with funding from The Utah Govenor’s Office of Economic Development, Centers of Excellence. |