Help me understand color space

madelaide

Premium Member

« on: September 27, 2011, 22:18 »

#link

Hi,

I recently had my parents' 50th anniversary photos scanned. I was horrified when I saw the results, all images have a very strong bluish hue, among other problems (such as a noisy texture on darker areas - low dynamic range, I suppose).

I noticed this when I used Windows' viewer (Win XP's Image and Fax Viewer), and when I opened the images at PSP, I got a message of "Color profile mismatch" and that it was replacing the embedded "Samsung Natural Color Pro" by the sRGB color profile of my working space. Then I opened the same image in Irfanview and the colors were much better. I try to edit them, I am not good at such strong corrections and the result was not quite good, especially skin tones.

Given that Windows viewer and PSP were not showing colors properly, I assumed there was a problem with the files, a problem that Irfanview was somehow able to address.

So I took the CD and the album back to the lab and talked to the guy who scanned them. He showed me the CD with Windows viewer in his PC (which has a Samsung monitor) and colors looked ok. Contrast was much stronger - a bit too strong for me - and there was little of the noise problem I had seen. He put the CD in another of their PCs, one that doesn't have a Samsung monitor, and the result was the same. He asked me to try at another computer.

I have a notebook, and the images also look bluish in it. The noise problem looks stronger, but that can be a monitor adjustment issue (I don't use the notebook for photography, so I really never bothered much about adjusting it).

I then went back to my PC and to PSP's settings and changed Color Management to "ignore embedded profiles", and then the images look better as I see in Irfanview - still with some problems, but not the bluish hue. From this point the edition is much easier and within my limited skills.

Now, what is wrong? Are the files wrong, should I insist with the lab? What have they done wrong?

I had some slides scanned with them in the past, and color was not much of an issue. But that was years ago, probably other scanner, computer and software.


	Ignore \| Logged

markrhiggins

« Reply #1 on: September 27, 2011, 23:00 »

#link

To understand colour space is complex. Each sensor or device records colour according to a gamut ( Colours it pick up). A defined set of colours is a colour space. The monitor and software have to interpret the colours and display them to a visual match. If for example you shoot in raw there is no colour space, it is all colours the sensor can record. In the raw converter we then open as a colour space and the software makes the conversion to the colour space. To make matters worse monitors have a different gamut and cannot display all the files colours (eg prophoto rgb)

If you shoot in raw for example and open in prophoto rgb it all looks ok. Save it as a different colour space and the colour will be way off and clipped. In PS you need to "convert" to the colour not "assign" the colour. This is similar to other software that opens the file "thinking" it is one colour space when in fact it is something different.

You need to convert the file to colour space it can interpret/display. I would convert them to prophoto in PS correct any problems and then convert to sRGB for display.


	Ignore \| Logged

markrhiggins

« Reply #2 on: September 27, 2011, 23:08 »

#link

below are some good links to read further. It is interesting that you may record a colour such as intense red on the sensor successfully but open in a color space that wipes it out. This may not be noticed because your monitor can not display it anyway. Better however to work with as many colors as possible even if they do not show on screen. Printers again have a different colour space and software needs to convert. What is not good is to throw away colous by choosing say sRGB that would have been printable.

http://www.luminous-landscape.com/tutorials/prophoto-rgb.shtml

http://www.cambridgeincolour.com/tutorials/RAW-file-format.htm

http://www.cambridgeincolour.com/tutorials/color-spaces.htm


	Ignore \| Logged

jsnover

Premium Member

« Reply #3 on: September 27, 2011, 23:40 »

#link

In addition to doing the general reading about the issue, it isn't clear to me whether or not you have a completely color managed environment.

What that means is that every device (printer, monitor, scanner) is calibrated so you have a profile for it that describes how it displays or reads color values. Every file has an embedded profile that says which color space the values in that file were recorded in - and that should ideally not be a device profile (like your monitor's profile) but Adobe RGB, ProPhoto, sRGB; one of the device independent color spaces.

Every piece of software that displays images should be set so that it reads color profiles and embeds them in files it saves (and Photoshop should be set to complain about missing profiles or mismatches with your default color space).

When you work this way, you'll know if you're opening something that has no embedded profile (which means you have to guess). What you should do, in your editor (PSP?) is assign (not convert) the Samsung profile to the images you got from the lab, and then convert them to the working space of sRGB.

Assigning a profile leaves the color values in the file alone and just tries interpreting them using the profile you assign. If you get that wrong - if you take an AdobeRGB image and open it as if it were sRGB, all the reds go dull and the flesh tones look awful - just try another profile until the image looks good.

Converting to a different profile (which you should only do once you've figured out which color space is closest to right) changes the color values in the file but the look on the screen shouldn't change at all. If you imagine a bright red color is displayed on the monitor, that might be 250 in sRGB and 254 in AdobeRGB (I'm just making up numbers). As long as the profile gets written into the image when you save it (which you must do after changing the color profile) the next person who opens it will see what you saw.

I'm assuming Irfanview was faithfully using the embedded Samsung profile, so the images looked good. I don't know anything about PSP, so I don't know how to make it honor the embedded profile on open and convert to sRGB after, but I'm sure you can find the help for that somewhere


	Ignore \| Logged

RacePhoto

« Reply #4 on: September 28, 2011, 01:09 »

#link

Oh No, this is going to be like White balance in two messages.

Your editing software might have settings like this, 1) No Color Management, 2) Adjust For Monitors, 3) Adjust for printing.

Plus you can have a color profile that's set for your screen or your printer...

Hard to tell from thousands of miles North. Wink

Have you asked the people who did the scans, what color profile they used. I think that would go a long way to solving the problem.

No matter what, it's not a loss, it's just and adjustment and once you get one right, all the rest should be the same. (note that was SHOULD...)


	Ignore \| Logged

microstockphoto.co.uk

« Reply #5 on: September 28, 2011, 07:04 »

#link

colour space is quite clear, as the set of all colour which can be represented

colour management instead is a way to apply corrections based on a device's profile supposedly to improve colour rendition, with the usual result of making things worse all the time by summing errors every time a new profile is added

this is my (mis)understanding of colour management


« Last Edit: September 28, 2011, 07:21 by microstockphoto.co.uk »	Ignore \| Logged

disorderly

« Reply #6 on: September 28, 2011, 08:36 »

#link

I hope this will help. In brief, a file format like JPEG gives you only 8 bits each for red, green and blue. That's 256 shades of each primary color, which isn't much. You might think that those 256 shades would be evenly distributed from black (0) to white (255), but they're not. It turns out we don't see small changes in intensity of color evenly; we can see changes in bright colors better than we can see changes in dim/dark colors. So the color space for red is an uneven distribution, to give us more subtle changes where we can see them and bigger differences where we can't.

Different color spaces try to optimize those 24 bits of color in different ways to give us the broadest range of colors. If you think of red, green and blue as three axes in a three dimensional space, a color space would be a stretched cube. The shape of the cube varies with each color space, but the goal is to have the biggest cube you can get with the best distribution of colors you can represent. But what's important is to note that a color value of 200,100,100 isn't necessarily twice as bright as one of 100,50,50. It's not a linear progression.

With that said, if you have an image that used one color space (say Adobe RGB) and an application that uses a different one (say sRGB), it will convert each color value to the closest equivalent. That should work pretty well, without much visible change. But if the image doesn't say what color space was used, or if the application doesn't know how to deal with color spaces, it'll likely get things wrong. A color of 200,100,100 might not show up as the sort of brick red the original image intended. It might be brighter, or more saturated, or redder (less green and blue), or some other subtle change.

And this is all separate from having all your devices color calibrated, so they show the colors they're supposed to show. I use a ColorMunki to calibrate my laptop's screen, so I'm more confident about the colors it displays. Most displays aren't all that accurate straight out of the box, and can change as they warm up or age. But it's important to keep the color information correct all through the process, both in hardware and software.


« Last Edit: September 28, 2011, 08:38 by disorderly »	Ignore \| Logged

digitalexpression

« Reply #7 on: September 28, 2011, 08:38 »

#link

Quote from: microstockphoto.co.uk on September 28, 2011, 07:04

No you got it right. Colour management is designed to standardize colour reproduction across a workflow. In other words you can calibrate your monitor and output device with the same profile used on your camera and you will see consistent colour on your monitor as you get from your printer and that will match (more or less) the colour from the images on your camera.

Unfortunately colour management is designed to work in a closed system. In other words, from YOUR camera to YOUR monitor to YOUR output device. If you apply a profile to an image that works within your calibrated system and then take that image to a different computer with a different calibration and therefore a different colour profile, the colour will be different.

Colour profile mismatches can be a mess. The first step is to strip out the colour profile that's currently embedded and apply your own. Personally, I don't use colour management at all. I routinely send images and files to clients and commercial printers and using colour management only mucks things up. Unless you work with the same clients and the same printer all the time and can calibrate every system along the whole path so that everyone involved is on the same profile, which never happens.


	Ignore \| Logged

microstockphoto.co.uk

« Reply #8 on: September 28, 2011, 08:46 »

#link

Quote from: digitalexpressionimages on September 28, 2011, 08:38

Personally, I don't use colour management at all.

I came to the same conclusion: without colour management, maybe colours are not perfect but at least not completely wrong


	Ignore \| Logged

gbalex

« Reply #9 on: September 28, 2011, 11:15 »

#link

Quote from: madelaide on September 27, 2011, 22:18

Your files seem to have been assigned a color profile to display files optimally on a Samsung monitor "Samsung Natural Color Pro", therefore when you open your image files the program that does it is looking for the Samsung - Natural Color Pro 1.0 .ICM file that Samsung installs on computers that use their monitors. The Samsung - Natural Color Pro 1.0 .ICM file has information that the proprietary monitor uses to display colors using Samsung hardware optimally. Who knows what .ICM profile Win XP's Image and Fax Viewer use instead when they can not locate the .com file, luckily Photoshop asks you.

It is easy to confuse assign .ICM profile with convert to color gamut which are two different animals.

I would ask the company to assign the AdobeRGB .iCM color profile to match the actual color gamut of the file. In the future ask them to scan to AdobeRGB which give you more color information than an sRGB file.


« Last Edit: September 28, 2011, 11:34 by gbalex »	Ignore \| Logged

madelaide

Premium Member

« Reply #10 on: September 28, 2011, 18:46 »

#link

So let me see if I get that. Correct me if I am wrong.

1) The ICM file for a monitor is supposed to convert information from any file into what is best for that monitor.

2) The lab has a ICM file for their Samsung monitors and it was embedded in the files with the scanned material. They should not have embedded it, because this is only useful with monitors compatible with that specific ICM file.

3) The real, "pure" scanned data is in the file and is what I see when I open it at Irfanview or with that ignore option that PSP has.

Now, when they used that profile, is it possible that they calibrate the scanner wrongly, so that what looks right in their monitor in fact generates the noise I see, if this can be called noise?

Or is it another issue, like monitor calibration? Or is it just a poor scan using a scanner without enough dynamic range?

I once has a monitor that I didn't notice it was already all the way up in brightness but still showed dark images, so when I scanned my slides I did the wrong calibration and the scans were actually washed out.


	Ignore \| Logged

digitalexpression

« Reply #11 on: September 29, 2011, 11:44 »

#link

Don't think that's noise. Remember that scanners see everything. If you touch a photo print with your fingers then scan it you will also scan your finger prints and they are quite noticeable. That can become incredibly annoying.

Your image looks like the scan picked up the texture of the photo paper. I've noticed that when I scan images I get better scans from prints that were done with glossy photo finishing than with matte because there's no "texture" to the print when glossy.

I could be completely wrong since I have no idea what kind of originals you had scanned but it looks to me like the scanner picked up more than you might have wanted. You may be able to reduce the effect in photoshop using the various noise filters like "dust and scratches" or "median".

edit: I should also note that I agree it doesn't look like a high quality scan. Scanners definitely are not all made equal and you get what you pay for so if the scanner the lab was using isn't high end it would certainly yield disappointing results.


« Last Edit: September 29, 2011, 11:53 by digitalexpressionimages »	Ignore \| Logged

gbalex

« Reply #12 on: September 29, 2011, 13:36 »

#link

Quote from: madelaide on September 28, 2011, 18:46

1. The images from your camera specify the color and brightness of those colors by mixing a certain amount of red, green and blue light. Cameras, monitors and printers each vary in how much red, green and blue they use to define red, green and blue, i.e. what frequency the manufacturer considers to be pure red, pure green and pure blue. The truth is, you can never quite cover the entire spectrum using the three primaries and each company tweaks their product to get the best results based on the components they used to build the device.

Monitor manufacturers, for example, will tweak the performance of a monitor by changing the characteristics of the (red, green and blue) elements. They do this to give you the greatest coverage of the color spectrum within the limits of their equipment. A cheaper monitor with a smaller color gamut range will generally use slightly different frequencies for red, green and blue than a more expensive model. To get truly accurate color on devices that define their color differently, you need ICM profiles that describe the characteristics of how an image or device reproduces color using the RGB primaries. You can think of an monitor ICM profile as a way of telling that monitor exactly how the primary colors should map to device independent color, i.e. a method for describing exactly what "color" you get when mixing the RGB primaries.

2. No they should not deliver a file to you that has an embeded ICM profile for their own color managed system. They could easily write a batch action to change it before delivering to the client.

3. Since you are not working in a color managed system I would simply change the ICM profile to Adobe RGB if you are working on a good monitor or sRGB if you are working with a cheaper monitor because they are usually only capable of displaying part of the sRGB color gamut.

I have no meaningful experience with scanners... it looks to me like the digitalexpressionimages is correct that the digital noise is generated by the scan. I would take one of the worst images to another agency to have them scan it, just to see if they come up with better results.


	Ignore \| Logged

markrhiggins

« Reply #13 on: September 29, 2011, 16:39 »

#link

samsung color pro is for monitor display management

http://natural-color-pro.software.informer.com/

God only knows the profile. Get them to supply in prophoto for best quality or sRGB for web display. Retail printing is best handled by aRGB


	Ignore \| Logged

madelaide

Premium Member

« Reply #14 on: September 29, 2011, 16:53 »

#link

I have to use these images for the discount coupons I have for photo albums. As I have already paid for them, and I already managed to get a postponement (they were due last Friday), I will see what I can do with these scans.

But I will return to the lab with the notebook and show them what the problem is. The way PSP has of bypassing the embedded profile works for most images, for the purpose of the albums at least (not a very high resolution), but some images are very weak indeed, and the histogram will show clipped highlights in some of them regardless of what monitor is used.

Thanks for your help.

But I have another question: when we use calibration software, do they really adjust monitor settings, or do they only create a "translation map" like an ICM file?


	Ignore \| Logged

markrhiggins

« Reply #15 on: September 29, 2011, 18:41 »

#link

A simple version is - for monitors spyders etc measure the colour put out by the monitors at points. It compares that that it should be. eg is 125,125,125 of red green blue showing as grey? It then builds a profile for the graphics card to display it correctly. So if it appeared to green it tells the graphics cards to reduce green etc. All these calibrations are in the graphics card profiling so it knows how to display the correct colour. Likewise for printers a card has preset colours and these are compared to output and software adjustments are made. NManufacturers' ICC profiles know how to adjust to different gamuts and colour space. Individual differences between units require profiling with measurement tools. Sme paper ink combinations etc may require new profiles to be meausred.

Calibration by eye is near impossible . It may be that your monitor displays light grey a bit green and dark grey a bit red. All the adjustments are easiest to make via the graphics display card adjustment rather than trying to mess with the monitor.


	Ignore \| Logged

Pages: [1]

« previous next »

MicrostockGroup

Microstockgroup Sponsors

Topic: Help me understand color space

Microstockgroup Sponsors

Microstock Poll Results