This is a splinter thread from the Superplex topic:
http://www.avsforum.com/avs-vb/showt...&pagenumber=36
Quote:
Originally posted by CMRA
marcorsyscom,
Your reposting of the applicable screen shot sent a shockwave through the other thread.
Maybe you can help clear things up. Perfect blacks, I gather are registered at 0,0,0 and perfect whites as 255,255, and 255 for each RGB channel. Am I correct so far? (shows what I know about Photoshop)
By this, does this encompass the 256 step gray scale we read and hear so much about?
If so:
Can I assume 0 is blackest black and 255 is whitest white? Thats only a 256 wedge scale, correct.
CR, or contrast ratio, is defined as the range between the whitest white and the blackest blacks.
So:
IF a projector has a CR of 500:1, how does it equate to to a 256 grayscale?
What if a projector has a CR of 1000:1?
What if a projector has a CR of 2000:1?
The real question is if the scale goes from 0 to 255 wouldn't a CR of 300:1 be all that is needed to get the job done?
Please educate me/us. |
Whoa, I'm a student, not an educator! I barely have a layman's understanding of the "256" numbers, and I'm the wrong person to ask about the science behind the specs and numbers. But I'm willing to start with some simple rambling and sit back to learn the real truth from those who really know...
I think "256" is practically arbitrary. Especially since we're more interested in the high/low ends (and the ratio between the two), instead of the number of slices (256) between white and black.
My understanding is that a digital camera's Auto White Balance (AWB) looks for gray and then divides the steps from the middle (gray) to the top (white) and bottom (black). So in a really bright scene, half of the color steps can be used in a small range to depict the subtle shades of white, without "clipping" (which would otherwise happen if the camera just said "everything above this brightness I'll call 'white'", so a lot of subtle shades and detail would end up getting called the same value of "white").
And if you didn't have AWB, then a dark scene would only have half the bits available to describe "dark", which means several shades of dark would get grouped together into the same color band, which would "crush" the blacks.
But back to "256"... A higher number of color bands used to describe the levels of gray (or in our case, levels of red green and blue) would be better, but 256 is usually accepted as "close enough for human eyes".
Consider the extremes: If you divide the range from black to white into only 10 steps, it's easy to differentiate between 20% black and 30% black. But let's say you divide the range between black and white into a million shades of gray. Could you tell the difference between gray#500000 and gray#500001? Probably not. That's why a JPG with 16.7 million colors looks better than a GIF with 256 colors.
But if you use 256 steps, you approach the reasonable ability to barely tell the difference between gray#125 and gray#127. So PSP uses a range of 256 steps.
Digital cameras do a great job of finding the brightest white it senses, and the darkest black it senses, then finding a balanced gray and dividing the steps to depict the shades and color.
But your question keyed on contrast ratio...
For contrast ratio, let's consider the low end and high end separately. CRT tube projectors do a great job at the low end (dark) because they can project black by just turning their electron beam OFF, thus not illuminating the phosphor, thus no light is generated.
But a bulb projector (LCD/DLP) has a harder time showing black because it's hard to stop light from a high wattage bulb with just a pixel. However, LCDs have an easier time with the high end (brighter whites). But even bulbs have their limits on brightness, so when you divide the brightest white (for LCD, let's say 1000 lumens), by the darkest black (let's say 2 lumens), you get 1000:2 which is the same as a 500:1 contrast ratio (all high:low values are only notional).
But CRTs aren't as bright, so let's say their brightest white is 250 lumens. However, CRTs can make great blacks (by turning the tube phosphor energy OFF), so let's call their black 0.01 lumens, so the CRT's contrast ratio could be 250:0.01 or 25000:1.
But black on a screen can only be as black as the ambient light. So if ambient light is 1 lumen, it moots the 0.01 lumen black of the CRT, so the effective CRT CR becomes 250:1. Conversely, LCD CR is less affected by ambient light.
And now back to how I propose any of this relates to the white and black values I measured in PaintShopPro from our screenshots...
This is not an attempt to debunk screenshots, but I think AWB and auto-exposure settings in digital cameras can compensate for whites that aren't really white and blacks that aren't really black.
The digicam looks for the brightest pixel it sees and calls it "white". Then it looks for the darkest pixel it sees and calls it "black". Then it does AWB and decides how to divvy up the steps between black and white to best depict the subtle shades and details without clipping the bright parts or crushing the dark parts.
My PaintShopPro (PSP. aka a poor man's Photoshop) was only measuring what the digicam had already adjusted for and recorded as a JPG... it found the brightest parts of the screenshot and saw that the camera stored it as "white", so PSP reported "255" (white), even though the screen was really a light shade of gray.
Then PSP saw the darkest part of the screenshot that the digicam called "black" (even though there was ambient, reflected, and bleed light from a very high wattage bulb in the room), so PSP dutifully reported the darkest pixels of the screenshot as 0,0,0 (black).
Phew. Sorry to ramble. I honestly look forward to better explanations...
-Clarence
There are a few other posts, but they're starting to get lost in the 37 pages of the original superplex discussion, so I'm going to start a new thread here.
My intent is to try to learn how most of the variables can be adequately accounted for in order to produce a screenshot as "accurately" as possible.
This is not a "debunking" thread!
