 |
Author |
Topic: Retinex (Read 520 times) |
|
Jerronimo
|
Retinex
« on: Nov 6th, 2003, 6:12am » |
|
http://www.cis.rit.edu/~sdlpci/Software/p5/retinex/
It demonstrates Edwin Land's process to generate (mostly) fullcolor images using just two primaries instead of the standard three (red, green, blue).
Basically, on every other line, just the red component is displayed. On the other lines, the Green component is displayed in the red, green and blue channels.
It's a process that probably never will be used for Image transmission, but it demonstrates how much our minds mangle incoming visual images into things we recognize, whether it be shapes or colors. Even with the inaccurate blue, and red tint, you still interpret the colors decently.
|
|
|
|
|
mrb
|
Re: Retinex
« Reply #1 on: Nov 6th, 2003, 8:38am » |
|
this is a vastly interesting aspect of signal transmission and processing -- i wonder what an equivalent function would be like with a sound file? thanks for sharing
mike
|
|
|
|
|
REAS
|
Re: Retinex
« Reply #2 on: Nov 6th, 2003, 8:39am » |
|
(mostly) is the key. 
very interesting. i wasn't aware of this particular phoenomena.
|
|
|
|
|
Jerronimo
|
Re: Retinex
« Reply #3 on: Nov 6th, 2003, 3:40pm » |
|
Yeah. it's pretty neat.
Your eye/cortex does a lot to make things appear the correct colors under various lighting conditions. Light a room with a colored light bulb, and you can still tell whcih things are white, and which are blue... Our brains are pretty darn amazing.
|
|
|
|
|
benelek
 
|
Re: Retinex
« Reply #4 on: Nov 7th, 2003, 6:14am » |
|
i'm not sure if it's the equivalent, but many lossy encoding types for sound files use perceptual coding (including MP3 and Ogg Vorbis).
basically, a whole lot of details are removed from sound that the human ear/mind and computer speakers either ignore or don't usually pick up. soft sounds are ignored when much louder sounds occur, the whole frequency range is narrowed to a maximum of about 15kHz, and fidelity is lowered within less sensitive frequency bands using power-level equations. All this and more, to squeeze a 30MB song into 3MB
|
|
|
|
|
Jerronimo
|
Re: Retinex
« Reply #5 on: Nov 7th, 2003, 2:41pm » |
|
Actually, one of the big things with perceptual audio compression is that the sound is broken up into frequency bands, then if there is a loud part in one band, the bands around it are not stored, since you can't discern those sounds anyway.
|
|
|
|
|
|
