Couble of questions about color...

[Moglex]

You do seem to be intentionally argumentative.

I do believe that that accusation could fit yourself as well as anyone.

Instead of simply saying that you believe that one definition is more appropriate in photographic terms you decided to be deliberately antagonistic by stating that my definition was wrong. And that without knowing the context in which the OP had seen the term.

There are at least three possible definitions (given that people are not always absolutely precise about the way the phrase things). The OP now knows them all.

 

[prodigy2k7]

lol crayons are old school, i use dry erase

 

[Senor Hound]

I didn't mean to start a fight. I promise!!! Please don't get angry at each other, you both are so helpful in your advice!

Here's the deal. To make sRGB you really need three channels of numbers from 0-255. That requires 8 bits per channel. So on the RGB spectrum, 8 bits x 3 channels = 24 bits. Yet, my Windows XP computer calls it 32-bit color, and there is no 24-bit selection (I would pick it if I could). I didn't know why this was.

I'd love to clarify even more, but I can't. After reading the following posts, I am somewhat confused myself.

 

[Garbz]

Very simple answer actually now that I know what you are talking about. 32bit colour is 24bit colour + an 8bit alpha channel. This is used internally in windows by all video cards that support alpha channels.

This allows each program to render on the screen a full 24bit scale, along with 256 distinct levels of transparency used by the windows GUI to pass to the videocard for hardware accelerated rendering of transparent windows (like when you drag a window and you can see through it). I did see something on wikipedia a while ago let me look.

/EDIT: lol it took me ages to find the most obvious wikipedia article ever :er: http://en.wikipedia.org/wiki/Truecolor

 

[Moglex]

/EDIT: lol it took me ages to find the most obvious wikipedia article ever :er: http://en.wikipedia.org/wiki/Truecolor

Please note the following from that article.

In a framebuffer, an alpha channel is meaningless, but 32-bit truecolor is often used because it makes pixel addressing easier. Addressing an array of 24-bit pixels requires multiplication by three, which is more computationally expensive than multiplication by four, which can be computed by bit shifting.

Which was what I originally said.

Of course the other meanings described by Bifurcator and HelenB remain correct as well.

 

[Helen B]

A thousand apologies. My misunderstanding.

Best,
Helen

 

[Bifurcator]

I didn't mean to start a fight. I promise!!! Please don't get angry at each other, you both are so helpful in your advice!

Here's the deal. To make sRGB you really need three channels of numbers from 0-255. That requires 8 bits per channel. So on the RGB spectrum, 8 bits x 3 channels = 24 bits. Yet, my Windows XP computer calls it 32-bit color, and there is no 24-bit selection (I would pick it if I could). I didn't know why this was.

I'd love to clarify even more, but I can't. After reading the following posts, I am somewhat confused myself.

It's a blatant error on Microsoft's part. (Anyone wanting to prove they are a moron please say: "it's nothing of the kind" and then make a bunch of unrelated out of context references to unrelated terms. Hehehehe... Oh, and don't forget to take the credit belonging to others.)
But the "simple answer" is as Garbz, Helen, or I provided in our respective "simple answer" sections.


What's still confusing you though? If it's the alpha channel thing just think of it as an embedded B&W (gray actually) layer used as sort of a cookie cutter layer for see-through parts of an image or graphic.

 

[Senor Hound]

It's a blatant error on Microsoft's part. But the "simple answer" is as Garbz, Helen, or I provided in our respective "simple answer" sections.


What's still confusing you though? If it's the alpha channel thing just think of it as an embedded B&W (gray actually) layer used as sort of a cookie cutter layer for see-through parts of an image or graphic.

So its almost like it controls the saturation? Sorry I'm not getting this. I'm really confused.

 

[Bifurcator]

I guess you could use the transparency layer to control saturation (in conjunction with a BG image). In PS terms I think they call that a mask.

The RGB+A embedded alpha format was "invented" for games and stuff. Think cookie-cutter. Like this:

It later gained wide use in digital compositing for stills and video (and of course GUI graphics like in MS Windows). Garbz also used the wrong term when he said "32bit colour is..." but I think he was defining what MS meant when they themselves mistakenly added the word color (colour) instead of "file". It's actually less of a nit-pick than it seems as there really IS something called 32 bit Color - and it's 32 bits per-pixel color or 96 bit file/image. As long as you're aware of the error it shouldn't present a problem. People who know what they're talking about Garbs definitely included, will usually specify around the common confusion that MS and Truevision Inc. caused by saying 32bpp or RGBA or something.

Here's some more:
http://www.am-guide.com/TinCan/AM_&_Alpha.htm
http://www.ypoart.com/tutorials/alpha/fundamentals.php
ftp://ftp.alvyray.com/Acrobat/7_Alpha.pdf
http://www.keithp.com/~keithp/porterduff/p253-porter.pdf
http://nicolaescher.com/tutorials/alpha-channel-primer.php

 

[Senor Hound]

Okay, so that allows for layering of color data... Which means its possible for one pixel to have two different sets of data, one being the base layer with full alpha, and one being a top layer with maybe only partial alpha... Hmmmm...

This is somewhat confusing. I guess it makes sense, but its hard to wrap my brain around at 3 in the morning.

 

[Bifurcator]

I can understand that. When I was 1st introduced to it I thought I understood it (but didn't) as being a 2 color padded B&W channel (or layer) used only for clip-mapping. When Adobe added "Layers" to PS I initially thought to myself - What a stupid use of clip-maps! LOL!

 

[Helen B]

The thing that caught me out, and possibly you, is that your original question appears to be about two very different things - one is capture bit depth and file formats and the other is the temporary storage of the display data.

In simple terms: for speed of access the information about the colour of each pixel on the display screen is stored in a particular way. It is very convenient to store the data in 'double words' - blocks that are 32 bits long - even if there aren't 32 bits of data to store. This means that the memory address of any pixel is very easy to find, and hence to alter. It is a storage method that trades compactness off for increased speed. 24 bits of data may be stored in a slot that has room for 32.

With an image file on a disk or other long-term storage - something that is saved - there is less need for this kind of speed and more requirement for compactness. Therefore if you only have 24 bits of data, you may only store 24 bits of data. When the file is opened it may be temporarily stored in a different way, with other priorities.

That doesn't mean that the file data will be stored with a total disregard of double word boundaries - typically the row data may start at the beginning of a double word. That's what happens with Windows bitmaps, for example: If you have three channels of 8-bit data you can store each block of pixel data in 24 bits of disk space as one option, or you can store four channels (one of which may be unused but available) in 32 bits of disk space. With both options the row data has to start at the beginning of a double word, so there could be unused file space at the end of each row.

Now that 32 bits per channel per pixel is in fairly common use there is plenty of room for confusion if 3-channels of 8-bit stored in 32 bits per pixel is also called '32 bit colour'. 32 bits per channel per pixel permits the storage of images in a reference frame that has a high dynamic range.

Best,
Helen

PS Have your other questions been answered adequately?

 

[Moglex]

It's a blatant error on Microsoft's part.

Actually, it's nothing of the kind.

If you look at my post (#9) where all three possible meanings are first laid out together, you can see that none of the candidates has a perfect claim to be the best.

1) A method of storing 8bits/channel colour data in Windows and other graphical operating systems to optimise processing speeds.

Not perfect because there is only 24 bits of colour information stored.

2) Normal 8 bit/ch colour with the addition of an extra channel (alpha channel) that is used in blending operations.

Not perfect because there is only 24 bits of colour information stored.

3) Very high accuracy 32 bit/ch colour depth.

Not perfect because there are actually 96 bits of colour information stored.

But the "simple answer" is as Garbz, Helen, or I provided in our respective "simple answer" sections.

And thank you all for confirming what I originally said in post #9. :mrgreen:

 

[Senor Hound]

Thank you, everyone. I appreciate all of the answers.