Archived posting to the Leica Users Group, 2012/03/22
[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]On Mar 21, 2012 at 11:41 PM -0700, Henning Wulff wrote: >Not quite. Boosting in software emphasizes noise. So it's better to >have two green pixels, so that the higher value of green required >doesn't require boosting the green 'one ISO value' up. This is definitely an issue as Henning points out. However, there are disadvantages other than noise to an equal distribution of sensor site colors, as the Bayer patent points out. If you did an equal distribution, say a stripe of R, followed by a stripe of B, followed by a stripe of G, you run into a couple problems. First is your horizontal resolution isn't the same as your vertical resolution. Second, our sensitivity to resolution in R and B is poor enough that you are wasting file size on things we can't see. Third, and possibly most important, is that the apparent resolution would go down, because instead of sampling luminosity (green) every other pixel, we only sample it every third pixel. To demonstrate how important luminosity and not color is to us seeing resolution, you can try this. Take a color photo into Photoshop and convert it into Lab mode. Apply a gaussian blur of say 4 pixels to the a and b channels. Notice how the picture doesn't change that much. Then apply the same blur to the L channel and watch how it goes to hell. Perhaps an easier way to do the same thing is watch a DVD. DVDs only have only have one color pixel for every four luminosity pixels (arranged in a square). So while the image has a resolution of 720x480 (in the US), the color resolution is 360x240. I'm pretty sure jpegs do the similar thing.