The Difference Between Layers and Channels

I'd like to clear up an all too common source of confusion for new Photoshop users (and apparently some old ones as well) here on the forums. That is the difference between layers and channels. The two words seem to be used somewhat interchangeably in much of the tutorial information that is generated here. Why this is, I simply cannot fathom, as they have absolutely nothing to do with each other.

(By the way, I said Photoshop out of habit. This is equally applicable to PSP, GIMP, or whatever your raster editor of choice happens to be, as well as to Photoshop.)

If I see one more tutorial that says "go to the channels tab and create a new layer," I think I'm gonna puke. You can't create a layer on the Channels Palette any more that you could create a channel from the Layers Palette. When you're on the Channels Palette, you create a new CHANNEL, not a new layer, and vise versa.

Telling someone to make a layer on the Channels Palette is like telling them to bake a cake in the refrigerator. Sure, the oven and the fridge look a little bit similar; they're both cubical, they both have a door, and they're both for putting food into before it's time to eat it, but they're obviously not the same things. Well, the Layers Palette and the Channels Palette aren't any more similar to eachother than are the oven and the fridge. They look kind of similar, but that's as far as it goes.

Also, I constantly see the term "Alpha layer" floating around. If anyone's been using that term, please cease and desist immediately. There's no such thing as an "Alpha layer." It's an Alpha CHANNEL. Telling someone to make an "Alpha layer" may as well be telling them to make a square circle. It simply can't be done, doesn't exist, impossible.

Let me offer some definitive explanations of what layers and channels are. Everyone should be using the proper terminology. There's simply no reason not to. I can't even begin to imagine how confusing it must be for new people trying to learn this stuff when so much so called "instructional" material constantly uses the wrong wordings.


Layers

Layers are elements of an image that are stacked on top of each other so that you can work on one without affecting the others. Think of them kind of like the way hand-drawn cartoons are made. It's all done in layers. Let's say Mickey Mouse and Donald Duck are gonna walk down the street together. Well, Mickey would be drawn on one sheet of acetate, Donald on another, and the street and background would be on a third. That way, both Mickey and Donald can both be moved around the scene independently, without affecting each other or the background. Because the sheets are made of transparent material, you can stack them on top of each other and it appears to be one single image with Mickey and Donald standing on the same street together.

Layers in Photoshop work exactly the same way. Let's stick with the same example of Mickey and Donald on the street. You've got a background layer on which you should paint the street and whatever buildings, landscape, etc. you want in the scene. On top of that you'd add a new layer on which you'd paint Mickey, and then you'd add a third layer on which you'd paint Donald. This way you can do whatever you want to any one element, and it won't affect the others. You could change the background from a street to a tropical beach, and Mickey & Donald would remain standing right where they are. You could grow Mickey a hundred feet tall, cut Donald's head off, whatever, and none of those changes to each individual element will affect the others in any way.

It's worthy of note that the example I just gave was extremely simplistic. It's not uncommon at all for a professional quality image to contain dozens, or even hundreds, of layers. For example, I just read a post by a skin maker here on the forums who said his typical skins have about 300 layers in them.

Layers allow you to easily change the composition of an image by moving them around, changing the order in which they're stacked, applying special effects to them, etc.



Channels

Channels are visual data maps that govern certain functions or attributes of an image. For color screen images, 3 such data maps (one each to govern red, green, and blue) are necessary to tell the computer how to create the image. Each channel is made up of pixels, the same amount of pixels as the image itself has. Each pixel in a channel controls a corresponding pixel in the image. Each specific attribute is controlled by a specific channel, pixel by pixel, with numerical values ranging from zero to 255. Those numerical values are represented visually as shades of gray, ranging from black (zero) to white (255).

This all seems a little complicated, but it's actually much simpler than it sounds. It will all make sense after a brief explanation of what's called Color Space, which is the term for how primary colors are combined to form an image. Images designed to be shown on a color screen exist in what's called RGB Color Space, meaning they are comprised of three colors, red, green, and blue. The relative brightness of each of these primary colors in each pixel determines each pixel's actual color. For example, a pixel comprised of maximum values of red and blue without any green would appear to be bright purple. (Everyone knows red & blue together make purple, right?) The computer determines the brightness of a primary color in mathematical values ranging from, you guessed it, zero to 255.

Now, remember I said 3 channels were necessary to tell the computer how to create a color image on the screen? Well, those channels are named Red, Green, and Blue. As their names would indicate, each one governs the amount of one of the primary colors present in an image, pixel by pixel. So, to create the example I just mentioned of a purple pixel, that pixel would be white in the Red Channel, white in the Blue Channel, and black in the Green Channel. Remember, in a channel, black means zero and white means maximum. The pixel therefore was given the maximum possible amount of red by the Red Channel, the maximum possible amount of blue by the Blue Channel, and no green by the Green Channel. Now apply that same concept to every pixel in the image and it's easy to see how the three channels together control the coloring of the whole picture.

So what's an Alpha Channel? Well, remember I said channels control image attributes? Those attributes don't have to be just color. Transparency is an attribute as well. Images with transparency have, in addition to the three primary color channels, a fourth channel called Alpha, which represents not color but opacity. In the Alpha Channel, black represents zero opacity (complete transparency), white represents maximum opacity, and shades of gray represent anything in between (translucency). The darker the gray any particular pixel is in the Alpha Channel, the more transparent that pixel will be in the image. The lighter the gray, the more opaque it will be.

That's about as far as channels go for SL purposes, but that's not as far as they go altogether. Since channels are just data maps, the data they contain can be used for all sorts of purposes. They can be used to govern the bumpiness of an image, the shininess, the liquidity, whatever you can think of. All SL currently allows for image attributes are color and transparency, but other applications can use channels for all kinds of things. I suspect one day SL will as well, but for now, color and transparency is all we need to discuss.


As you can see, layers and channels are not synonymous at all. They have absolutely nothing to do with each other except that they are both present in an image, just like the oven and the fridge are both present in a kitchen. Hopefully this post has cleared up any confusion between the two. If anyone has any questions, let me know.

Or a fifth, or sixth, or hundredth. You can have as many alpha channels in an image as you want while you work, though for SL transparency it's best to export with only one combined alpha channel.

A very nice use for them is as stored selections. White is fully selected, black is not selected, and grey is translucent partial selection. There's a button at the bottom of the Channels Palette that creates an alpha channel from the current selection (Windows / PS). Then later control-click an alpha channel to use it as a selection. Helpful!

That's all I have to add. Good stuff, Chosen.

Thanks for this explanation. I've been curious about alpha channels for quite awhile.

I'm a little unclear about a fine point. If an image has multiple layers, is it correct to say each layer has red, green and blue channels? If so, does each layer then have its own alpha channel as well, and can you then adjust the transparency of each layer just as you can adjust the colors of each layer?

Thanks Again,
Savannah

You can adjust the transparency of each layer depending on what sort of alpha channel ... file ... you have.

For instance, Photoshop CS needs a seperate alpha channel for ALL the layers. It encompasses them ALL. This means that it doesn't matter what you do with the individual layer in question in terms of transparency.

However, in Photoshop 7 (I think?) you can just adjust the transparency for each layer and save it as is. (This is the MUCH EASIER OPTION) Please take a look at this thread if you have Photoshop CS: /109/8b/48537/1.html. Whatever is transparent in Photoshop will be transparent in SL, providing you save the .targa file as 32 bits and not 24.

The answer to this is, I think, a little trickier than it seems.

Each layer can, and indeed does have it's own RGB channels (in an RGB image) and other layers, such as alpha.

The final image ALSO has them - in effect in anything except .psd format it's flattened to a single layer (at least for the formats we use in SL). PS does some funky maths to work out the effects of the layers and their channels into that final image but the upshot of it seems to be that it assumes any layer without a particular channel has that one at full white and then paints darker colours on to the indicated points, equal to the darkest colour present in that channel in any of the layers.

The practical upshot, at least when I tested before answering this, is that you make a background layer for the cut out of your clothing shape you don't have to have alpha channels on any other layer for it to work properly. If a bit later on you want a specifically shaped cut-out you can add an alpha channel on that layer, your final image will still be transparent in all the right places. [I bow to Chosen's superior knowledge of how PS does its thing.]

One thing I want to know: photoshop displays transparancy as the old white and grey checkerbox behind the image, but when you have an alpha channel the only way I have found to view the transparancy is to toggle both it and the RGB channels and then there's no checkerbox, just an arbitrary background of one color and one opacity of your choosing.

It's probably nitpicky, but I'm used to gauging transparancy by checkerbox, is there any way to just have the alpha channel on and applied to the image so you can work without worrying about switching back and forth all the time?

My attempt at an explanation (only applies to photoshop)

Savannah - your short answer is "yes".

The long answer is:

Each layer in a Photoshop file has one channel per colour value, and an optional alpha channel. RGB files use 3 colour values per pixel, CMYK use 4, Indexed use 1, etc etc
Keeping the discussion relevant to SL, I'll only consider RGB.
So each layer has at least 3 channels :- Red, Green and Blue (duh!)
If the layer has what Photoshop calls a "layer mask" then that is the optional Alpha channel for the layer. (This mask can be raster or vector, doesn't matter for this discussion though)

In the "Layers" palette, the thumbnail on the left of each layer's bar is the combined R, G and B image - and the Alpha channel AKA layer mask will be just to the right, if there is one.

In the "Channels" palette, the top bar (RGB) is the composite of your image.
This is not a true channel. Think of it as a conveniant way of getting back to "normal" editing after you have been editing a specific channel.
Below that are the R, G and B channels. These are also composites, taken by combining each colour's channel from all the visible layers. Don't worry about them for now, plenty of time to learn them later.
Below the four channel bars we have just discussed is... nothing... yet.

As you create channels, they will appear here. The really important thing to remember here is that the channels are not linked to any particular layer at all. They are completely independant of layers. In fact, many applications don't know they exist.

For SL work - use one of these. This is what SL is going to interpret as the texture's transparency. The layer masks are great for mixing layers together, but should always add up to fully opaque. What I mean by this is - you could, for example, have a solid blue layer. Above that you have a solid green layer. The green is opaque, so your image is all green. Now we add a layer mask to the green layer, and fill the mask with a white-to-black gradient. Remember that this mask is an Alpha channel - so where it is white, the green is opaque. Where it is black the green is transparent. What we seen now is a gradient from green to blue. If you upload this to SL, that's what you'll get there too.

Now delete the blue layer. In Photoshop you now have a gradient from green to "checker-board". If you export this to TGA and upload to SL, and you'll get a green-to-white gradient texture. Or save yourself the L$10, and just open the TGA in Photoshop Photoshop "fills in" any less-than-opaque areas with a white matte.

To get a green-to-transparent texture, you would rather drop the layer mask, leaving a solid green layer, and then add a solitary alpha channel that cotained the white-to-black gradient.

---------------------
This answer is now 4 times longer than I was anticipating; has no pictures; and is no doubt longer than your hand.
So I'm going home now.
byeeeee

... after I answer Emma's question
I usually have a solid matte colour right at the bottom of my layers. Choose a colour that matches the edges of your artwork to prevent halos if your alpha channels are anti-aliased.

While I work this layer is switched off, and I use the checkerbox as you do.
Just before uploading, I dump my old Alpha channel, ctrl-click any layer thumbnail (or layer mask thumbnails for those layers that have them) to load it's alpha as a selection.
Then ctrl-shift-click all the other visible layers, to add their alphas to the selection.
Then create a single Alpha channel from the selection.
Switch on the the matte layer - save - save as - upload - bingo.

Obviously, as you edit further, you will only need to re-do this when the alpha values change. If you are just tintintg colours, or over-painting etc, you can keep the old alpha.

k - NOW I'm going home

I've got responses to many things people have asked about and said so far in this thread. Portions of this post will correct various bits of incorrect information people have stated either definiteively or assumptively. If you are one of the people whose statements are to be corrected, please don't be offended. My purpose here is only to make sure the right information goes out. In that light, it matters not who anyone is, only what has been said. Nothing here is personal, so please don't take it as such.

I realize that most people wouldn't, but there are a few people on this forum who tend to get hideously bent out of shape whenever their theories are challenged, so I figured I should address that in advance. So, for the record, if your name is mentioned in a quote header, it's only because citing where a quote came from is always the proper thing to do. It's not an attack or a mark of shame or ridicule in any way if something you've said gets corrected. Please take it for what it is and be an adult about it.

Okay, let's get started.



Every picture is comprised of channels, and that includes layers, so technically each layer has its own set of color channels. However, the RGB channels Photoshop shows you are not layer-specific. They reflect everything that is visible in the image. So, to see what the RGB channels of an individual layer look like, turn all the other layers off.


Again, the answer is technically yes, but functionally not necessarily. Every time you erase part of a layer, paint something on it, change the overall opacity, etc., you are changing the transparency data map for that layer. However, that datamap is not directly accessable for channel-style editing as it is for the image as a whole. If you want that functionality, you can add it by using a layer mask. See "Masking Layers" in the Photoshop help files for more details. I'll probably write a tutorial on maks some time in the future, but I haven't done it yet.


With respect, I STRONGLY recommend against doing what that thread says. If you have Photosho CS, or any version of Photoshop other than 7.0, leave it exactly as is. If you have 7.0, get the free 7.0.1 patch from adobe.com. DO NOT change your TGA save process file to act like that of 7.0. Photoshop 7.0 was Adobe's one and only attempt at automated alpha channel creation for TGA images, and it didn't work. It was an idea that sounded really good in concept, but turned out to be really, really, really bad in practice. All it did was eliminate the level of control that manual channel editing allows, and it made every single image prone to the white halo effect without any way to get rid of it. Here's an exerpt from a previous thread where I talked abut this in some detail:






Please, you're doing exactly what this thread was designed to prevent, which is you're mixing up the words "layer" and "channel". Think about what you said. Each layer has its own layers? No disprespect intended, but please realize that doesn't make sense. The correct thing to say would have been each layer has its own channels.


I'm not sure what you mean here. If you're talking channels, they behave exactly the same way in all formats, including PSD. If you're talking layers, there are other formats that use them besides just PSD. For example, TIFF uses layers the same way PSD does, and GIF uses them for animation purposes.


Not sure where you're getting this. Layers are completely transparent until you paint something on them. To put that in alpha channel terms, that would make them "black" until you paint on them, and then they're "gray" or "white" in the areas you painted, depending on the opacity at which you painted. They remain "black" in all unpainted areas.

Understand I'm speaking metaphorically here. As I said earlier, while the actual internal transparency maps for individual layers do exist prgramatically, they are not user accessable as visual data, so there isn't really any black, gray, or white in a litreral sense. The closest functional equivilent is a layer mask, but that's unrelated to any data contained in the layer itself.


I'm not sure what you mean here. When you make the alpha channel, it will affect the image as a whole, all in one step. I'm not following you when you say things like "you don't have to have alpha channels on any other layer," and "add an alpha channel on that layer." Alpha channels are not assigned to specific layers. They globally affect the entire image.

As I've said a couple of times now, masks are similar to alpha channels in their functionality, but unlike channels, masks can be layer-specific. I'm assuming that's not what you were talking about though, since making a mask is a very deliberate process, and it's unlikely someone would talk about them without mentioning them by name. So what exactly did you mean?


EDIT: Emma's question appeared while I was typing. Ben's answer was pretty good, but let me offer some of my own thoughts as well.


You can work exactly the way you're used to, alpha or no alpha. Just make sure your background layer is either turned off or that it's been changed to a regular layer. To make it regular, double click on its name in the channels palette.

The alpha channel only needs to come into play upon output to TGA. What I usually do is work against the checkerboard, just like you do. My lowest layer is always a dark color to prevent haloing. I keep it turned off until it's time for output. My final steps before output are to turn the background layer on, create the alpha channel, and then apply the alpha as a mask to the background so I can see how the outputted image will look. If everything looks right, I delete the mask, output to TGA, and it's all good.

Great post and this helps fill in some of the missing parts in my quest to understand image editing. I am interested in more help in removing the white halo in textures I have created, it is very frustrating to me

So if I read this right, all I would need to do is make the color of the background black or a good dark color to match the actual artwork I will be working on, add a new layer to perform the actual artwork on, then apply the alpha channel to the background and this will remove the white halo and instead make it a darker color halo, perhaps so close to the actual artwork it will not show up? Is this the thing that is called bleeding?

i've nothing important to contribute, except a book plug - don't know that it's allowed; jeska, you may delete this if i've violated forum rules here...

"real world photoshop" is the book to have, if anyone wants to learn how photoshop -really- works. those "wow" books just don't do much explaining -why- their tricks work the way they do... i always felt like an automatic-transmission driver trying to figure out a stickshift. then i found "rwp", and well, that was the real "wow" book for me.

highly recommended - and if you're on a budget, look at the used amazon prices on the older editions, those still have all the basic important info.

Corrected the post. It was a silly mistake, sorry.


What I was saying, or trying to say, in a half-baked and possibly backwards fashion was that channels as we see them are a composite. It you click so you can only see one layer then the channel information you see is the channel information for that layer. You can, from that make an alpha channel that is clearly associated with just one layer. If you then do things on other layers PS combines the layers and channels information, and that alpha channel attached to just one layer will affect the whole image.

Whilst I agree with your analysis of the other channels being black until painted on I'm not sure I do for alpha channels, although in CS2 it does appear that way when you make an alpha channel. Perhaps it's my technical confusion showing through. The way I think of PS combining the layers and the resulting composite channels information suggests that for an alpha channel it assumes 0 transparency (full white) unless told otherwise in a particular place. Imagine making a simple T-shirt with a design on it. You can put your T-shirt template on one layer. Remove all the other layers like shape outlines etc. Add an alpha channel so the arms, neck etc. are the shape you want/need. We've got one layer and 4 channels (R, G, B, alpha). In a new file make your design for the T-shirt, just RGB channels. Copy and paste this onto the T-shirt template. The overall alpha applies, and the new layer shows fully opaque. Ergo, to my thinking, the alpha channel associated with that design layer is white, not black. Maybe that's wrong and the channel is copied directly onto all the layers?