Learn something new every day More Info... by email
Wavelet noise is a pioneering function that is applied in three-dimensional computer graphics. When graphic artists create three-dimensional textures, they face problems like loss of detail and aliasing with existing noise functions. Wavelet noise functions overcome these problems and allow texture creators to make more realistic textures. It was developed by Tony DeRose and Robert L. Cook at Pixar Animation Studios, who published a paper on the topic.
Creating three-dimensional textures or visual elements in computer graphics involves using noise functions. Realistic-seeming smoke, clouds, and fire are a challenge to create within a three-dimensional, computer-generated environment. The most popular noise function used is called Perlin noise and was developed by Ken Perlin. First used to create spectacular visual effects in the motion picture Tron in 1982, Perlin won an Academy Award from the Academy of Motion Picture Arts and Sciences for his contribution.
The Perlin noise function is an industry standard and allows visual effects artists to create natural-looking synthetic textures. It suffers from a few very specific problems, though. At high spatial frequencies, the image loses a great deal of detail and looks artificial, especially when distant objects like the horizon are represented. Aliasing is also a problem; it results in distortions in the image or texture. The loss of detail versus aliasing are critical problems that become very apparent when three-dimensional noise is used to texture a two-dimensional surface within the graphic industry.
Attenuating the bands to reduce aliasing effects leads to other undesirable consequences. When magnifying a scene, the texture details are seen later when compared to the geometry, and both elements don't seem to be tied together. The texture also fades in a way that appears totally artificial. A method called spare convolution tried to address these problems but did not completely solve them.
Wavelet noise functions avoid all these issues. They are an alternative to using Perlin noise functions to create natural-looking textures in three-dimensional graphics. Quite easy to implement, it allows visual effects creators to use three-dimensional wavelet noise to texture a two-dimensional surface quite well because it is excellently band limited.
It permits orthogonal bands and allows the shape of the spectrum to be controlled more finely. The noise generated is completely band limited, allowing a great amount of detail to be created with very little aliasing. The wavelet noise function allows a noise function with a hard frequency cutoff to be created quite easily.