What gamma is LAB color space?

sRGB = 2.2
Adobe RGB = 2.2
Prophoto RGB = 1.8

L*A*B* = ?

???!

LAB is not a colour space, in the same way that RGB and CMYK are not colour spaces. ProPhotoRGB, AdobeRGB, and sRGB are all ways of defining how you move from the three Red, Green, and Blue, primary colour values that make up the image data, to a physical colour produced by any specific output device. Effectively it is defining just what shade of orange pink or blue any particular set of three values should have. The thing is that we can describe colour in many different ways, RGB values being one of them.

We can also describe colour in an image as a subtractive set of the complementary colours, Cyan, Magenta, and Yellow. In this case 0,0,0 is white, and adding a value will apply more ink to the paper, making the colour darker. Because this method of describing colour is normally used in commercial offset printing, which results in muddy looking dark colours, it is usual to also add some black ink, which is given the value K, so that we end up with four values.

There are still other ways to describe colour, and one that was used in analogue electronic images is to describe the grey scale brightness, or Luminance, and then two values to describe the colour intensity along two different axes, this is called chrominance. Again there are many specific device/application variations of how this is implemented. This system is how old analogue colour TV implements colour, but still allows for viewing the image in black and white on older TV's that have no hardware for decoding the colour information. Internally JPEG files also convert the image data to a luminance/chrominance space, as this way you can remove a lot of the data concerning the colour, without drastically altering the perception of the decoded information. Because JPEG requires the conversion to this different form of representation, you will ALWAYS get some conversion artifacts when saving, or resaving a JPEG file, even at maximum quality settings.

Lab colour in Ps works in just the same way. It can be very useful for doing some tasks, noise reduction would be one example, as it allows you to separate the brightness of the image from the chrominance. In the case of Lab you get the gray scale luminance values in the L channel. The a channel gives a (based on a quick inspection of the channels in Ps) Magenta to Cyan axis, while b gives Yellow to Blue. It is well worth while opening an image in Ps, and converting to Lab colour, then playing with the visibility of the three channels. You will very quickly understand how JPEG can get such high levels of data compression, with very few artifacts. Almost all of the images fine detail is stored in the L channel, so simply converting to Lab, then halving the bit depth of the a and b channels will reduce your data storage needs by a third, for very little perceived loss of image quality.

Alan

So if I import an untagged photo of a 21-step grayscale wedge into Photoshop ("untagged" meaning there is no color space assigned to the image), then what number value does the middle gray patch have, if any? If I send the photo to the printer, doesn't the printer need a number value so it can know how much ink to lay down? All the instructions for printing color charts and step wedges say to print the file without a color space assigned. How is that even possible? The printer driver doesnt have eyeballs. It cant "see" the step wedge. It can only digest numbers. But no color space means no numbers.

Well you don't want to be using Lab colour for that, you need to convert to RGB, or greyscale. If the image is a true gray scale image it will have a single luminosity value for each pixel, in either 8 or 16 bits, although using 32 bit would also be possible. When using grey scale you usually have the choice of using a gamma of 1.8 or 2.2. A true grey scale image obviously cannot have a colour space, since it has no colour data.

The other option is to use an RGB image, where for something like a grey scale wedge you would simply have R=G=B, which in the absence of a colour space profile should show as grey. The issue with this is that with printers especially passing any triplet that is R=G=B many not actually result in a print that is totally neutral in tone colour wise.

When it comes to printing a test print to profile your printer the best way to do that is to simply turn off colour management in BOTH Ps and the printer driver. In other words in each case set it so that the colour management is being done by the other system. What will happen then is that the image values will be passed to the printer without any alteration to get them to match a specific shade of colour.

In an 8 bit image a midtone grey will have the following RGB values 128, 128, 128. This will be mid grey, but it will be the output devices idea of mid grey. If it is mixing equal quantities of three inks to achieve this it may turn out that the result is a little pink, or maybe yellow or any other colour. By using a working colour space like sRGB what you effectively do is say that these three values should be "this" exact shade of colour. The output device will also know that compared to the standard set of colours, it needs to adjust the inks a little to get an exact match. So R=G=B=128 might require the printer to receive 124, 129, 132 to end up with an exactly neutral grey.

What the colour profile does is specify exactly what shade of red green or blue the maximum value represents, and also just how much each change in value will change the colour by. An RGB image without an assigned colour profile still has three numbers associated with it, but it lacks any means of knowing exactly what red, green, or blue really represents.

The working profile sRGB was designed to match the actual output from the average crt monitor from around 1995. It was done that way so that if you didn't include the profile, the image would still be reproduced relatively accurately on screen. The other working profiles, most usually AdobeRGB and ProPhotoRGB have more red reds, greener greens, and bluer blues. To do this they also have to have the steps between the colours be wider. ProPhotoRGB should only be used with 16 bit colour depth, and I would generally say that AdobeRGB would be better in 16 bit than 8 bit per colour.

When your software does colour management it basically looks at the standard colours of the working profile, the values you use for editing, and which will pretty much always result in a neutral tone being R=G=B, and converting those numbers to the numbers to get the correct colour output from the output device. To do this a set of test values have to be passed to the output device, and the device must produce the colour that those values represent.

Once you have produced the output you then need to have a device that can accurately measure the colours that the device actually produced, if you are calibrating a monitor or other screen type device you will do that measurement one colour at a time. For a printer though you will print swatches of each colour on a page, and measure them all at the end. What you then get is a table that allows you to convert your RGB values from the working standard profile, to the values to match that specific output device.

If the work you are doing is very critical for colour accuracy, such as when working with things like corporate logos, or other products, you will need to keep all of your devices correctly profiled, so that you see the same exact colour on each device. That is pretty simple on a monitor, but printers are harder. The problem is that every batch of ink that the ink manufacturer produces will be ever so slightly different. Then the way that ink interacts with the paper will differ for every production batch of paper, let alone paper type and brand. For most general work using a generic profile for a particular ink type and printer, along with the brand and type of paper will produce excellent results, it isn't good enough for some super critical work.

In those situations you would want to profile your actual device, for a particular lot of ink and paper. Change any of the ink cartridges, or the batch of paper and you would need to recalibrate. This is though working at the extreme limits with very picky clients. If you are working with OEM ink then the supplied generic profiles supplied by the paper manufacturer to match a printer will be more than adequate for almost everybody. Profiling you own printer is more important when using non OEM inksets, although for most it will only really be necessary to profile for the ink, and the paper brand/type. I would not worry too much about doing a new one for every cartridge change or paper batch.

Alan

Lab is NOT equal to RGB! You can either be RGB Exclusive OR Lab Exclusive OR CMYK. Of course you can convert from working in one domain to another, and there are obviously numerical equations that will convert from one to the other. For what you are trying to do you need to completely forget about Lab or CMYK colours.

When working in RGB, without colour management, we have to appreciate that we are working with three undefined points that represent Red, Green, and Blue. When we look at colours we usually use one of those graphs that look like a curved triangle of colours, where the axes of the graph represent the wavelengths of the light needed for that colour. Without colour management you don't know exactly where on that graph your maximum red, Green, or Blue value will fall. So to correct this we use colour management, and this effectively has two phases. The first is a working profile, the main three being sRGB, AdobeRGB, and ProPhotoRGB. These standard profiles define where on that colour graph the three corners of the RGB triangle will lie. So it defines exactly what colour we should see when we set each colour channel to it's maximum value. We now have a definition for the colour.

Regardless of where the corners of the triangles lie on the colour graph, the number of steps along each side is always the same, in the case of an 8 bit image 256 values. So if the corners are further apart you have to have a bigger gap between them. sRGB is the smallest triangle, and has a small step size, even when using only 8 bits per colour. AdobeRGB is a bit larger, while ProPhotoRGB is the largest, and should only be used with 16 bit values.

The Gamma curve associated with any standard colour profile is just like any other curve that you can apply in the Ps Curves tool. Basically for any input value x the output value will be some function of x, usually represented by f(x). The function is any mathematical function that you care to use. It might be linear, such as x+2 or x×5, or it could be nonlinear, such as x squared. IIRC a Gamma curve is usually x raised to the power of the curve value, so x^.18 or x^2.2. This means that small values are only increased by a small amount, while larger values are raised by a larger amount.

This is usually only done to the linear output from a camera's sensor. This is because humans don't see in the same way that sensors do, and the gamma curve corrects this deficiency. To move from one Gamma curve to another one is difficult, as you would first have to convert the data back to a linear format, and then apply the curve again. The problem arises that if you initially used Gamma 2.2 for example, many of the higher values would have been clipped in the process. So you cannot reverse the process, since you don't actually know the correct value to use. If you want to then use a lower value Gamma you would have clipped fewer values. So Ideally you should only set the gamma of an image once, during RAW processing, and then not change it again.

You can apply a gamma curve to a grayscale image that uses a single value to represent brightness, just as easily as to an RGB image that is simply using R=G=B, the math is unaffected. 123^2.2, 123^2.2, 123^2.2 is completely equivalent to 123^2.2, where the second two values are simply implied when outputting to a colour hardware device.

The second half of colour management is the output device. Every output device has a different capability as far as what colours it can output. So if you send the device the maximum value for red, it will output it's maximum redness, and the same for green and blue. These will be different to the values that are used to define a working colour space. In general most output devices are limited to hues that are less intense than the working space, so although your working space allows more intense colours, you have no way to output them on the device. If we go back to the colour graph the device will also have a triangle associated with it, with corners in different places. Now in each of these triangles, just like the working spaces, the number of steps from one corner to the next is always the same, so if we want the most saturated red we would send the device the code 255,0,0.

The problem is that this is not going to match the hue that is specified by our working profile. So where the colour required by the working profile is within the capabilities of the output device all we need to know is that to match a particular hue you need to change the number from the working value, to the matching device value. This can be done in a number of ways, and sometimes more than one method might be used in a single conversion. The most accurate way is to build a complete lookup table, you look up the pixels value for the working profile, it might be 26, 154, 200 and the LUT will then send 23, 154, 220 to the device so that you see the correct colour in the final output. As well as a LUT you can simply apply a curve to each channel, which again could be linear, or exponential, it will often depend on the level of accuracy required. LUT's are great, since you can have completely arbitrary changes. Or you can have a mix of a curve of simple to convert tones, and use a LUT for any difficult parts.

As I mentioned when you send a grayscale image to a device it will only have the single value, but the computer system is programmed with the information that this should be interpreted as R=G=B. You don't actually need to know any more about the image, since generally regardless of the device when we use 0 it implies that we want the darkest tone possible, i.e. the blackest black, while 255 is the brightest possible white, and the tones in between will always be relative to those absolute white and black points of the device. On a device that simply uses black ink, and uses dithering to represent grey tones, that is all you need to worry about.

When though you are using a colour device, you might have a problem in that sending the device 127,127,127 it might not actually produce a pure grey, often on many printers they end up with a slight green tint. In that case you would want to send the three values that will result in a completely neutral tone. In the case of a green tint for example you may end up with 127,120,127 removing some of the green colour in the image. This is why it is necessary to use an output profile, even with a grayscale image.

Most colour managed image editing programs will allow you to do the colour management conversion in one of two places, although it is possible to do it in neither, or both! You can either do the conversion in the program itself, in which case the program will perform the conversion, and send the corrected values to the printer driver. This is very useful where you have a basic device, or at least a device with a basic driver, that doesn't have a lot of options. Or you can send the values directly to the driver, and have that do the conversion for you. Generally you will have an option in the imaging software, and the device driver that will allow you to turn the colour management functionality on and off.

The one situation that you must ALWAYS avoid is having colour management enabled in both the program, and the device driver. If you do that the program that you are using will convert the values to be correct on the output device, and then the device driver will apply the same correction again, to the already corrected values.

The other situation you can get is where you have management turned OFF in both the program and the device driver. In that case the values will be sent directly to the device without any conversion being applied. In this situation it won't matter what the working profile was, if you send 255,0,0 to the device, it will produce the most red output it can. This is how you produce a test print for building your own colour profile for the device, since it allows you to measure what is output for any particular set of input values. The more input values you use to build this profile the more accurate it will be. If the output device works in 8 bit that would mean that you would have to output (print) 16777216 colour swatches, and then measure the actual colour hue of all of them!

For your monochrome printing you should be fine sending it a grayscale image using a Gamma 2.2 curve from your RAW conversion. Since it is a colour device you will still need to use colour management, and if printing from Ps I would usually do the colour management in Ps, not in the driver. From what you are saying if you are going to send it a colour image, or an RGB image with R=G=B then you should simply use AdobeRGB as your working colour profile, since the printer allows for sending those values without any conversion. You should still use colour management, but the values are simply passed straight through.

I think you are worrying too much about this really. If you have a wide gamut monitor, and this printer it is now worth your while actually using AdobeRGB as your colour profile when working in Ps with RGB images. At least if you are starting with a RAW file. I would also try to keep the image in 16 bit/channel too, since AdobeRGB is quite a bit wider than sRGB. If you only have existing RGB images with the sRGB working profile, then it is not worth converting them to AdobeRGB, since you will lose some colour fidelity in the conversion, since the image will still only have colours in the sRGB gamut.

Finally even if you did all of your work correctly, very often you don't have many/any colours outside of the sRGB gamut in a natural scene. Mostly you need to have manmade subjects with bright mostly primary synthetic pigments to get any colours outside of sRGB.

Alan