Lately I have been finishing my underwater images in Photoshop, as Lightroom does not offer the amount of color correction control I waned.

However, reimporting my corrected images into Lightroom, the black point RGB numbers were not  consistent with my Photoshop RGB numbers.

My normal workflow is:-

1. Import my RAW underwater images into Lightroom and optimize them as far as possible.
2. Open the image into Photoshop with the Lightroom adjustments.
3. Do a final colour correction in Photoshop using a Curves Adjustment layer and  RGB numbers. Using “Color Correction by the numbers” tends to give the image more punch, with blacks set to 12:12:12 (5%  black) and whites to 243:243:243 (5% white). (This technique is explained in detail in Photoshop CS4 – Color Correction by Taz Telly on Lynda.com.)
4. Save the Photoshop modified image with layer modifications in TIF format.
5. Import the modified TIF file back into Lightroom.
6. Knowing the final RGB numbers in the modified TIF file, do a final adjustment to the original RAW file in Lightroom, to get the RAW and TIF images as close as possible.

When I set my blacks and whites to 5% in Photoshop and imported the images back into Lightroom, I noticed the blacks were not 5% black.

I was stumped on this for a while until I did a bit of experimenting to figure out what was going on.

Figure 1 – RGB1998 Test pattern.

Creating the test pattern in Figure 1 using a RGB1998 colour profile, the RGB 12:12:12 (5% black) in Photoshop measured 2.2% black in Lightroom.

Likewise the RGB 243:243:243 (5% white) in Photoshop measured 4.6% white in Lightroom.

What is Happening?

To fit in as many colors as possible, Lightroom uses the Melissa RGB color space (named after one of the Lightroom team members), which approximates the ProPhoto RGB profile, and is larger than the RGB1998 and sRGB profiles.

When my Lightroom image is exported to Photoshop, the image is converted to my RGB1998 working color space, which alters the black point.

This means:

1. If I have an image in Lightroom that I want to adjust for 5% black in the working RGB1998 color space, I need to set the black points in lightroom to 2.2%.
2. If I have an image in Lightroom that I want to adjust for 5% white in the working RGB1998 color space, I can use the Lightroom setting as 95.4% is close enough to what I want.

]]> http://www.carlchapman.com/workflow/colour-management/lightroom-rgb-percentage-vs-photoshop-rgb-numbers/feed/ 0 http://www.carlchapman.com/workflow/colour-management/firefox-fixes-color-rendering-issue/ http://www.carlchapman.com/workflow/colour-management/firefox-fixes-color-rendering-issue/#comments Tue, 04 Aug 2009 01:46:20 +0000 Carl http://www.carlchapman.com/blog/?p=474 Mozilla have fixed a problem with Firefox browser version 3.5.2 rendering ICC color profiles incorrectly that seem to happen on some particular monitors (e.g dell, nec) and monitors with wide gamut.

• Initial Problem description.
• Fix Details.

Going to the browser configuration menu (see my previous post Firefox V3 Browser Colour Management), I notice the gfx.color_management.enabled element has been removed.  Now only gfx.color_management_rendering_intent and gfx.color_management.mode are listed.

Using the default setting of management.mode =1 from the upgrade (new browser default is 2), I get the following result:-

Image 1

Notice how version 3 (right) displayed the color profiles correctly, but now version 3.5  (left) is not displaying e-sRGB and YCC-RGB correctly. The GBR and Adobe 1998 segments are displayed correctly.

I have looked around the web and apparently the reason for this change is that Firefox 3.5 no longer supports ICC Version 4.

I have ended up configuring my browser with the options:-

• gfx.color_management.mode = 2  (Color management applied only to tagged images.)
• gfx.color_management_rendering_intent = -1  (Use embedded intent)

For more information on these settings see the Mozilla developer forum – ICC color correction in Firefox.

I am traveling at the moment, so do not have the benefit of my large screen, but am running my colour profiled laptop using a Windows XP Pro platform.

I checked a favorite page on CNET that has a number of test images. The Browser displays the images as being color managed, but it is like the profiles are being displayed incorrectly.

I viewed the CNET images in three different browsers. Left browser is Firefox with color management enabled, center is Chrome, and right is Safari 4.

Both Firefox and Safari display as colour managed browsers, but the Safari browser is displaying the profiles of the top and bottom image differently, where Firefox is displaying the profiles correctly.

This is quite confusing as some tests  on the David Harradine  Image test page, and directly viewing images from the drive in safari4 are OK.

If I view my images on iStockPhoto they are displayed OK in Safari4, however if I view my images on Alamy or Photoshelter they are not displaying correctly in Safari 4, where they used to display OK in Safari3.

The images on all of these test sites display correctly in my color managed Firefox browser.

It seems like one of the methods for determining colour profiles has been removed in Safari 4.

I am not exactly sure what the error is, but would be interested in anyones comments.

Magenta Cast in Raw Images

Canon has released a firmware upgrade to Version 1.0.2.0.

According to Canon this upgrade is to fix the following issue:-

If RAW images are captured by continuous shooting under the ISO 1600 setting, abnormal data is recorded and a magenta cast appears in the second and subsequent RAW images. This phenomenon also occurs in the second and subsequent recorded RAW images captured by single shooting, but only if the LCD monitor display mode is set to the OFF position.

This phenomenon cannot be confirmed when images are played back on the camera’s LCD monitor. It can only be confirmed if image processing software (such as the Digital Photo Professional software bundled with the product) is used to develop RAW images on a PC.

For more information visit:

• Power shot G10 Firmware update info (Canon Japan)
• Canon Australia Support note

Lines Appear in Captured Images

According the Canon Australia website:-

In certain circumstances, lines may appear in images captured with some PowerShot G10 units.

1. This phenomenon may occur irrespective of the camera settings.
2. The occurrence of this phenomenon depends on the subject and may not occur every time.
3. Images that exhibit this phenomenon cannot be fixed.

This problem seems to only effect specific batches of the G10. Refer to the Support note on the Canon Australia website to determine if your camera needs to be returned for repair.

For more information visit:

• Lines Appear in Captured Images – Canon support site.
• Canon Issues Service Notice Over “Lines” in Images from  Some G10 Cameras – PDN

Collapsed Underwater Housing O-ring

I was checking the yellow o-ring on my Canon WP-DC28 underwater housing before a dive yesterday and noticed what looked like a pinch mark over a 3 mm segment.

This housing was brought in Dec 2008 and has done around 10 dives.  I always inspect the o-ring before and after each dive, and it is stored in a separate plastic bag between dives, so I know it has not been pinched, and expect the o-ring has collapsed around where it was joined. (I have owned 3 other underwater housings for Canon cameras and have not seen this before.)

After some chasing I discovered the part number for the problem o-ring is CY4-4517. I have ordered some new o-rings from the spares department, and sent the faulty item to Canon support for further inspection.

Links:-

• Underwater housing preparation and Care – Canon
• Underwater housing preparation and Care Demo – Canon

The Problem

I was recently scanning some images using my CanoscanFS4000US slide scanner (Canonscan FS4000US ICC profile).  After scanning the images and assigning the custom profile, a number of slides were exhibiting saturated highlights.

Figure 1 – Saturated highlights

Figure 1 shows the ear of a Koala bear. You can see an area of red pixels around the bright centre of the ear.

I had not seen this in the past and it had me confused for a while, as I thought using a custom profile was supposed to correct this.

With a little investigation I was able to determine my problem was the scanner settings.

Scanner Settings

The Canoscan FS4000US scanner has a number of user configurable settings such as Exposure, Dust Removal, Focus, Colour Matching and Gamma.

I set the gamma to 2.2 and turned off all the auto settings.

Figure 2 – Exposure setting options.

The Auto Gain checkbox was left checked (ON), thinking this was an additional function to the exposure setting.

I then scanned one of my Wolf Faust (www.targets.coloraid.de) IT8 Provia 35mm transparency targets.

Figure 3 – IT8 target with Auto Gain ON.

Checking the scan results, a number of the light colored boxes are blown out/bright white (Figure 3).  Deselecting the Auto Gain checkbox (OFF), and using a zero exposure (EV0), rescanning the target produced the same result.

As EV0 was causing saturation, I decided to try EV-1. (Unfortunately the manual exposure setting for this scanner only works in +/- 1 steps.)

Figure 4 – IT8 target with Auto Gain OFF and EV -1.

Rescanning the IT8 target with Auto Gain OFF and EV-1, Figure 4 was a bit darker as expected, but the lighter squares showed some colour and detail. I also checked the blacks in the bottom row to make sure each was still distinguishable and they were not saturating.

Rescanning the target with Auto Gain ON and EV-1, I got the same blown out image as Figure 3. From this I can only deduce that the Auto Gain function overrides and disables any manual exposure setting.

Assigning the Profiles

I created separate ICC profiles from the IT8 target with Auto Gain ON and OFF, and assigned them to copies of the scanned image in Photoshop.

Figure 5 – Basic scanned image.

Figure 5 shows the scanned results for the same transparency. The left image was scanned with Auto Gain ON and EV-1. The right image is scanned with Auto Gain OFF and EV-1, and is darker as expected. At this point there are no signs of highlight saturation.

Figure 6 – Applying the ICC profiles.

The ICC profile generated with Auto Gain ON  and EV-1 was assigned to the left image in Figure 6, and the profile generated with Auto Gain OFF and  EV-1 assigned to the right image. Notice that in assigning the profiles to each image, the mid tones and colour casts where also automatically corrected. (There is still a slight red colour cast, but this can be improved later in the workflow process.)

Figure 7 – Close up view of highlights.

Zooming in on the ear in Figure 7, the area of red pixels formed from saturation can be seen around the bright area in the left image, but the image on the right is showing no saturation and significantly more detail.

Figure 8 – 3D view of compressed and uncompressed ICC profile.

The effect of Auto Gain compression on the pixels can also be demonstrated in viewing the ICC profile with the Chromix ColorThink software in Figure 8. The solid profile is the same for Auto Gain ON EV0, Auto Gain OFF EV0, and Auto Gain ON EV-1. The wireframe shows the profile for Auto Gain OFF EV-1.

Conclusion

A couple of important things I have learned in this little experiment are:-

1. Turn off ALL your scanner Auto settings when doing a scan. Using manual settings can make image correction easier and more accurate.
2. Be careful when profiling a scanner. Look for blown out highlights or saturated blacks in your target scans. You may have to fine tune the manual settings to get the best results.

]]> http://www.carlchapman.com/workflow/colour-management/scanner-settings/feed/ 0 http://www.carlchapman.com/workflow/colour-management/canoscan-fs4000us/ http://www.carlchapman.com/workflow/colour-management/canoscan-fs4000us/#comments Wed, 24 Dec 2008 05:51:55 +0000 Carl http://blog.carlchapman.com/?p=162 I decided to get serious about profiling my Canoscan FS4000US slider scanner, due to having several years of slides from the period before my switch to digital.

My Canoscan FS4000US scanner is a couple of years old now (costing around $1500AUD when new), but provides high resolution images from slides and negatives.

For years I have wondered about what profiles to apply to my scanned slides. I tried hunting around on the web to see if there was some kind of boxed ICC profile available, but could not find anything.

For a while now my standard procedure has been to import the images from the scanner into Photoshop, assign the sRGB profile, and then convert to my default workspace profile of AdobeRGB. I have tried assigning AdobeRGB profiles directly to the scanned images, but it resulted in over saturated images.

Noticing my images were slightly washed out after conversion, I began to question how much color information has been distorted in the workflow process.

Producing the scanner ICC profile

The first step was to purchase a couple of IT-8 slide targets from Wolf Faust (www.targets.coloraid.de). They were well priced, supplied very promptly, and good quality. Wolf has lots of interesting stuff on his main site www.coloraid.de about color management.

To produce the profile, I then downloaded and purchased the Profile Mechanic software from Digital Light and Color, which made the process of generating a profile very easy, and gave me the ICC profile required.

The generated profile is specific to the individual scanner, settings used, and the type of film the images are shot on. Most of my work was on Fujichrome Velvia100 and Provia slide film. I generated profiles for both, and even though they are generated for my scanner specifically, I have made them available for people to play with.

• Velvia and Provia ICC Profiles ( 6.4 MB zip file) – Updated 10 Jan 09 for high accuracy and better highlights (Settings:- Auto gain off, exposure -1, Screen Gamma 2.2)

Using the ColorThink software from Chromix, and comparing the new profiles to  the AdobeRGB and sRGB profiles, the scanner profile was surprisingly large.

The profile comparison above is for a Provia transparency slide, scanned at a gamma of 2.2.

The real world result

Scanning a sunset image, I assigned the CanonscanFS4000US provia profile, converted it to AdodeRGB, and then used the curves option in Photoshop to bring up the highlights.

The curves screen shows the pixel input is distributed from 0 to 207.

Assigning a sRGB profile to a copy of the original image, and converting to AdobeRGB, the curves screen shows a gap in the area of darks and highlights.

Using the curves tool to bring up the darks and highlights, the input pixels were distributed between 10 and 222.

Therefore the dynamic range of pixels is about the same in both images.

A closer inspection of the curves graphs show more pixel information in the shadows region of the curve for the ICC profiled image. This produces a more natural and correct result compared to the original scanned image.

I am now in the lengthy process of going through my slides and reprocessing them using the correct scanner ICC profile.

I recently purchased the Canon Powershot G10 so I could shoot RAW from a compact platform, and take advantage of the extra bit depth.

I have been using Lightroom2.1 and CS3/Adobe Camera Raw (ACR) 4.5 for my processing workflow. Neither of these supported the G10 CR2 RAW format.

Note: – As I am writing this article, the Lightroom 2.2 update has just been made available to support the G10.

For ACR support of the G10 RAW format I have to update to CS4 with ACR5.2, and  am still in two minds as to if the upgrade it is worth the cost.

The Canon Digital Photo Professional software (v3.5)  supplied with the G10 exports to TIFF and JPEG, but not DNG.

After a bit of searching I found the latest version of Adobe DNG Converter (5.2) supports the G10 RAW format (as well as Canon 5D Mark 2,  Leica D-LUX-4 and Pansonic’s DMC-G1, DMC-FX150, DMC-FZ28, DMC-LX3 cameras).

Using the converter I am able to change my canon raw files to DNG format and then work with them in any version of Lightroom and ACR.

DNG Profiles

When I downloaded the DNG converter, I noticed it was around 35MB compared to the previous version at around 10MB.

A bit of research revealed that the reason for this extra size is that the converter also contains the new Adode camera profiles.

In the case of Canon cameras, these profiles are designed to emulate the rendering of the preset Picture Styles in the Canon DPP software. They act more like Lightroom presets rather than ICC profiles. This subject can be a little confusing, but visit the following links for more info:-

• Adobe Labs – DNG camera profiles
• Canon Digital Photo Professional (DPP) preset Picture Styles

Once installed, the DNG camera profiles can be found in in the Camera Calibration section of Lightroom and ACR.

The profiles are not the “be all and end all” of processing, but rather a good place to start, and can provide some punch to your images when starting out. Once you have selected the profile you want, then you can move through the normal workflow process of white balance, tone adjustments, sharpening, etc.

DNG Profile Editor

Another feature to come out of the camera profiles is the DNG Profile editor.

The Profile Editor software allows you to modify the present profiles and create news ones.

Even more interesting is the ability to use the Profile Editor software with a 24-Patch Color checker chart to create custom DNG profiles for different white balance situations.

This is an exciting bit of software, and I expect we will see more of this in the future.

For interest, I decided to compare the canned profile with the Spyder 2 colorimeter profile using the Chromix ColorThink software.

The 2D view of the profiles show the canned and measured profiles follow each other quite well, except in the area of blue to red transition. It also shows they are off-set slightly.

Investigating the 3D view (canned profile in solid, measured profile in wireframe), shows the reason for the offset is that the white point has a different placement in two profiles. Rotating the 3D view I  checked the black points and they located in the same place.

Conclusion:- The canned profile will probably do for the everyday computer user, but anyone serious about colour accuracy and management should be profiling their monitor.

Ideally I would have liked to purchase an Eizo screen, but they aren’t within my budget just yet, so I looked at various screens available around town.

Settled on a Samsung 2243BXW due to price, and the fact that Samsung is one of the four companies that make all of the LCD monitors in the world.

The thing that I like about the screen is the 300 cd/m2 brightness. This blows away my laptop, and interestingly this does not seem to be a specification typically listed with laptops.

The first thing I did on setup, was color calibrate the monitor using the Pantone Spyder2 to compare it with my HP laptop.  After some experimenting I discovered Windows does not support ICC profiles on multiple screens, unless separate video cards are used.

The other thing I noticed is that there is a very narrow viewing angle (around 5 degrees) for the correct colour viewing. Moving slightly to either side (or up and down) and a red hue begins to appear at the edges.

After calibrating the screen, I checked the ICC profile using Chromix ColorThink software. This software is brilliant for understanding color issues, allowing visual 2D and 3D rendering and comparison of color profiles.

Results of the calibration show the Samsung 2243BWX screen has a color profile that approximates sRGB. (The Eizo screens quote 95-98% of AdobeRGB1998).

The other result is that the laptop screen gamut is smaller than sRGB, and this explains why images viewed on a laptop in a non-color managed environment appear faded. (I have had a couple of friends confine this effect on different laptop brands.)

This raises an interesting question – If LCD monitors can produce sRGB, why do laptops have a smaller gamut, when they are effectively the same thing?

The conclusion – The Samsung 2243BWX will do me for now, but when I get the funds together, will be upgrading to an Eizo monitor.