3 ms·

Without dithering, 6 bit LCDs produce 191 colors. 8 bit LCDs produce 767 colors. If you want more than 767 colors you will be using dithering. I think where pe

by jws 17y ago

Without dithering, 6 bit LCDs produce 191 colors. 8 bit LCDs produce 767 colors. If you want more than 767 colors you will be using dithering. I think where people get confused here is the naive dithering where a group of 3 pixels (one of each color) are logically grouped together gives your the 252k and 16m numbers, but there is no reason that is the only desirable solution. Consider sub-pixel rendering of typography. This abandons the traditional groups of 3 pixels and uses alternatives to gain better horizontal spatial resolution. Consider a white, lowercase 'L'. Depending where it appears on my screen it could either be a single white pixel under the naive "groups of three" dithering, or a yellow pixel next to a blue pixel. The catch is that unless I show you the borders of the screen you can't tell which it is. Fortunately, I have a brand new 15" MacBook Pro under my fingers. Tonight when it gets dark I'll take a picture of a gradient, enlarge it beyond the capabilities of my eyes, and we will have an answer about what Apple does.

jws 17y ago
So far, on large areas I can see a difference between #404040 and both #3f3f3f and #414141 on the new 15" MacBook Pro. That rules out a naive 6 bit display. I think. The application of color profiles can could cause some regions of to have one-to-one mapping even if there aren't enough output codes available to cover the inputs, but I think with the 4:1 difference I can rule that out.
- jws 17y ago
  Here is a little web page I tossed up to help people screen LCDs. http://jim.studt.net/lcd-sleuth/ http://jim.studt.net/lcd-sleuth/ It will make a table with cells whose text is just slightly off of the background. You can stare are them and explore the limits of your visual system and your computer's display. (No one will be able to see all of the text on any statically adjusted display.) Additionally, on this MacBook Pro I looked at a greyscale gradient which I have used over the past and decided it was poorly distributed, so I ran through the color calibration (expert mode) in the Displays preferences pane. I ended up with a much better distributed gradient, but I introduced hue distortion so I didn't keep it. But there seems to be room for an improved profile.
duskwuff 17y ago
You are confused. "6 bit displays" have 6 bits of color data per channel, and hence can display 2^18 (262,144) colors without dithering. I'm not sure where you're getting 191 and 767 from.
- jws 17y ago
  I am in no way confused. The three color components of an LCD pixel are not in the same spatial location. Look closely and you will find three rectangular areas, one for each color. The only colors you can make without combining elements are at the 2^6 or 2^8 variants of red, green, and blue. I subtract two because (ok, I am confused, I only subtracted one before) because the blacks are all the same.
  - duskwuff 17y ago
    1. If you're seeing the RGB elements separately, you need to sit a little farther back from the screen. :) I can't pick them out individually at all (I can just barely see the pixel grid with the screen an inch from my face), so it's not particularly useful to think of color this way - because nobody sees it like that. 2. Even ignoring this, your numbers are still wrong. 2^6 = 64, and 2^8 = 256.
    - jws 17y ago
      Continuing our long, invisible thread… 2^6 = 64, but I have red, green, and blue pixels, so times three, but subtract two because all the blacks are the same. I suppose I'll need to write a small article on this with some pictures. In the process of researching this I was impressed with how useless macrophotography of a screen was. Small changes that are obvious to the human at a distance are invisible when looking at the individual r,g,b elements up close. I'm away from my camera that can output RAW data, maybe with it and some numerical image analysis I can tell if there is multipixel dithering going on, but my human eye isn't going to tell, even from magnified pictures.
      - duskwuff 17y ago

by jws 17y ago

jws 17y ago

So far, on large areas I can see a difference between #404040 and both #3f3f3f and #414141 on the new 15" MacBook Pro. That rules out a naive 6 bit display. I think. The application of color profiles can could cause some regions of to have one-to-one mapping even if there aren't enough output codes available to cover the inputs, but I think with the 4:1 difference I can rule that out.

jws 17y ago
Here is a little web page I tossed up to help people screen LCDs. http://jim.studt.net/lcd-sleuth/ http://jim.studt.net/lcd-sleuth/ It will make a table with cells whose text is just slightly off of the background. You can stare are them and explore the limits of your visual system and your computer's display. (No one will be able to see all of the text on any statically adjusted display.) Additionally, on this MacBook Pro I looked at a greyscale gradient which I have used over the past and decided it was poorly distributed, so I ran through the color calibration (expert mode) in the Displays preferences pane. I ended up with a much better distributed gradient, but I introduced hue distortion so I didn't keep it. But there seems to be room for an improved profile.

duskwuff 17y ago

You are confused. "6 bit displays" have 6 bits of color data per channel, and hence can display 2^18 (262,144) colors without dithering. I'm not sure where you're getting 191 and 767 from.

jws 17y ago
I am in no way confused. The three color components of an LCD pixel are not in the same spatial location. Look closely and you will find three rectangular areas, one for each color. The only colors you can make without combining elements are at the 2^6 or 2^8 variants of red, green, and blue. I subtract two because (ok, I am confused, I only subtracted one before) because the blacks are all the same.
- duskwuff 17y ago
  1. If you're seeing the RGB elements separately, you need to sit a little farther back from the screen. :) I can't pick them out individually at all (I can just barely see the pixel grid with the screen an inch from my face), so it's not particularly useful to think of color this way - because nobody sees it like that. 2. Even ignoring this, your numbers are still wrong. 2^6 = 64, and 2^8 = 256.
  - jws 17y ago
    Continuing our long, invisible thread… 2^6 = 64, but I have red, green, and blue pixels, so times three, but subtract two because all the blacks are the same. I suppose I'll need to write a small article on this with some pictures. In the process of researching this I was impressed with how useless macrophotography of a screen was. Small changes that are obvious to the human at a distance are invisible when looking at the individual r,g,b elements up close. I'm away from my camera that can output RAW data, maybe with it and some numerical image analysis I can tell if there is multipixel dithering going on, but my human eye isn't going to tell, even from magnified pictures.
    - duskwuff 17y ago