Page 1 of 1

Color Cast Removal. bug?

PostPosted: Mon Oct 14, 2002 9:11 am
by Gregory C
First, a beginner's question.

Why is Auto Color Cast Removal ignored when IT-8 calibration or ColorSync is selected for CMS:Scanner > Internal (thereby implementing a full color management workflow)? I don't understand the reasoning behind this.


Now for the bug.

Procedure:
  • In the General panel, set Pos./Neg. to Positive.
  • Click on the Options button and navigate to the CMS panel.
  • Select ColorSync for Scanner > Internal, and select a scanner profile in the Profile section.
  • Navigate to the Auto panel and set Color Cast Removal to any non-zero value.
  • Close the Options dialog.
  • Set Pos./Neg. to Negative.
Note that the Image Auto-adjust button is still colored and still contains the C in its centre. These indicate that the color management workflow is still in effect, and that any color cast will be retained.
  • Prescan an image.
  • Click on the Image Auto-adjustment button.
If you now bring up the Gradations dialog, the position of the Color Cast Removal slider will be at 0 indicating that Color Cast Removal is inactive. A look at the Channels Top Down histogram will further confirm that Color Cast Removal has not been applied.

When a color management workflow has been turned on for Scanner > Internal, Auto Color Cast Removal is turned off. This workflow is only affective for Positive film scans though. When Negative is selected, the Scanner > Internal method and the scanner profile are ignored by SilverFast. This switch is done automatically. It appears however that SilverFast is forgetting to activate Auto Color Cast Removal when the switch to Negative is made.

Is this a bug?

regards
Gregory

Re: SF Ai v6 ? Bug 12? Color Cast Removal. v6.0.1r23

PostPosted: Mon Oct 14, 2002 7:21 pm
by ilyons
[quote="Gregory C"]First, a beginner's question.

Why is [color=darkred]Auto Color Cast Removal[/color] ignored when IT-8 calibration or ColorSync is selected for CMS:Scanner > Internal (thereby implementing a full color management workflow)? I don't understand the reasoning behind this.


Now for the bug.

Procedure:[list][size=10][*]In the [color=darkred]General[/color] panel, set [color=darkred]Pos./Neg.[/color] to [color=olive]Positive[/color].
[*]Click on the [color=darkred]Options[/color] button and navigate to the [color=darkred]CMS[/color] panel.
[*]Select [color=olive]ColorSync[/color] for Scanner > Internal, and select a scanner profile in the Profile section.
[*]Navigate to the [color=darkred]Auto[/color] panel and set [color=olive]Color Cast Removal[/color] to any [color=olive][u]non-zero[/u][/color] value.
[*]Close the [color=darkred]Options[/color] dialog.

[*]Set [color=darkred]Pos./Neg.[/color] to [color=olive]Negative[/color].[/size][/list]Note that the Image Auto-adjust button is still colored and still contains the C in its centre. These indicate that the color management workflow is still in effect, and that any color cast will be retained.[list][size=10][*]Prescan an image.
[*]Click on the [color=olive]Image Auto-adjustment[/color] button.[/size][/list]If you now bring up the [color=darkblue]Gradations[/color] dialog, the position of the Color Cast Removal slider will be at [color=olive][u]0[/u][/color] indicating that Color Cast Removal is inactive. A look at the [color=darkblue]Channels Top Down[/color] histogram will further confirm that Color Cast Removal has not been applied.

When a color management workflow has been turned on for Scanner > Internal, Auto Color Cast Removal is turned off. This workflow is only affective for [color=olive]Positive[/color] film scans though. When [color=olive]Negative[/color] is selected, the Scanner > Internal method and the scanner profile are [u]ignored[/u] by SilverFast. This switch is done automatically. It appears however that SilverFast is forgetting to activate Auto Color Cast Removal when the switch to Negative is made.

Is this a bug?

regards
Gregory

[size=10]Amateur bug detective[/size] :D[/quote]







Gregory,

[quote]Why is [color=darkred]Auto Color Cast Removal[/color] ignored when IT-8 calibration or ColorSync is selected for CMS:Scanner > Internal (thereby implementing a full color management workflow)? I don't understand the reasoning behind this.[/quote]

Auto Colour Cast Removal is applied to highlight and Shadows NOT the midtones. It's assumed that a calibrated scanner will already have neutralised these end points and so any colour cast that may exist is a function of the actual image and/or film stock.

BTW: you use of HTML is a first order in pain-in-the-ass for folk to try and read when quoting - stop using it and MAYBE we can read the quote instead of flicking back and forth to the original! Have alook above to see CRAP we're trying to read!

PostPosted: Mon Oct 14, 2002 8:57 pm
by ilyons
Gregory,

Having copied and pasted your message into something more appropriate it's now possible to address your question.

[quote]

Now for the bug.

Procedure:.............................

..............

* Set Pos./Neg. to Negative.
Note that the Image Auto-adjust button is still colored and still contains the C in its centre. These indicate that the color management workflow is still in effect, and that any color cast will be retained.

[/quote]

You are correct in so far as the CM workflow for CCR is still functional (i.e. don't use) but the actual ICC profile and CM system is OFF and so no automatic COLOUR corrections are made by this tool.

CCR only works on Highlights and Shadows - keep that in mind!


[quote]

* Prescan an image.
* Click on the Image Auto-adjustment button.

If you now bring up the Gradations dialog, the position of the Color Cast Removal slider will be at 0 indicating that Color Cast Removal is inactive. A look at the Channels Top Down histogram will further confirm that Color Cast Removal has not been applied.

[/quote]


Actually it's the Histogram dialog that shows CCR and the display is correct.


[quote]

When a color management workflow has been turned on for Scanner > Internal, Auto Color Cast Removal is turned off. This workflow is only affective for Positive film scans though. When Negative is selected, the Scanner > Internal method and the scanner profile are ignored by SilverFast. This switch is done automatically. It appears however that SilverFast is forgetting to activate Auto Color Cast Removal when the switch to Negative is made.

[/quote]



When negatives are being scanned you don't want an ICC profile to be applied and you DON'T want CCR to be used either! The colour balance of negatives is AUTOMATICALLY optimised by NegaFIX. Try pressing the Expert button in NegaFIX and see the 3 channel histogram. The colour balance (and removal of the mask) of the image is altered by NegaFIX setting the end points for RGB channels. Look at the highlight values in the histogram, now move the Auto Tolerance slider and see that ONLY the highlight values change.

Now go to the the MAIN histogram and open it without touching any controls and look at the RGB channel values for highlight and shadow. NegaFIX will have made the first stab at getting the maximum dynamic range and also neutralising the endpoint colour cast. You can manually move the CCR slider and tame any residual cast in the highlights. However, I always think that you should aim to have achieved this in NegaFIX using the Auto Tolerance slider

Remember neither CCR or Negafix Auto Tolerance will have ANY impact on a midtone colour cast - that's why you have the NegaFIX profiles (curves) and all the other tools still at your disposal.



[quote]

Is this a bug?

[/quote]

I doubt it and if they fix it to match your expecations it'll be broken.

Ian

HTML? It's BBCode, a function of this BBS's architecture

PostPosted: Tue Oct 15, 2002 12:39 am
by Gregory C
BTW: you use of HTML is a first order in pain-in-the-ass for folk to try and read when quoting - stop using it and MAYBE we can read the quote instead of flicking back and forth to the original! Have alook above to see CRAP we're trying to read!

The 'HTML' is not HTML. It's BBCode, a function provided by the BBS php system, and a function which in many cases allows better expression of complex issues.

Your reply (and the replies of many other people) are difficult to read simply because you neglected to uncheck the 'Disable BBCode in this post' option when you wrote the reply.

If you find it cumbersome to uncheck the 'Disabled BBCode' option each time you post a reply, you can uncheck it in your profile and that will become the default for each message you create.

If you absolutely abhor the BBCode, then simply select and copy the text in the message you wish to quote before clicking on the Reply button. Then in the message form, click on the Quote button once (optional), paste the message text and click on the Quote button once more to close the tag (optional).


Your point about trying to read the quoted text while replying to it is a good point. It's a shame that the system doesn't provide split panes when replying.

The system should also perhaps automatically uncheck the 'Disable BBCode in this post' button whenever the Quote button is used because it's effect requires that BBCode be turned on.

regards
Gregory

PostPosted: Tue Oct 15, 2002 8:24 am
by ilyons
Gregory,

Thanks for the info on BBcode. I didn't know what it was for or what it even is. I have ZERO interest in most of the Bells and Whistles (CRAP) that is supposed to enhance our web experience. This forum and ALL those that use such enhancements only make for "MY" experience to be negative. Java, Flash and all the other crap fall into the same category - the sooner the whole damn lot gets junked the better!

Ian

My understanding of Calibration, Negafix, Color Cast Removal

PostPosted: Tue Oct 15, 2002 4:54 pm
by Gregory C
Here is my understand regarding Calibration, Negafix and Color Cast Removal.


Calibration

Calibration serves to match the colour data received from the scanner with known colour data. It does not perform any colour balancing, mask removal or color cast removal. It ensures that the scan data after digital mapping is as true a representation of the film as possible.

While some people may disagree, I believe that calibration should be independent of Negafix, producing accurate scan data which would include the orange mask if present. Negafix could then be applied to accurate calibrated scan data. If possible, I would prefer to use SF Ai to produce calibrated HDR scans of negative film, and then use SF HDR's Negafix to extract the orange mask and invert the image.

In my mind, turning on ColorSync for Scanner > Internal is equal to turning on IT-8 Calibration only if the corresponding scanner profile is accurate. If the scanner profile is a generic profile for the scanner model, turning on ColorSync does not represent an accurate CMS workflow; just an automated workflow. Beggars (such as myself) cannot be choosers though and using a generic profile is probably better than not using any profile at all.
    (In case any other Super Coolscan 4000 ED owners are interested, I was able to produce SilverFast-recognised generic icc profiles for the scanner from the Nikon-provided profiles for 14-bit and 8-bit scanning using a HEX editor to replace the 4-character profile type code. If anyone would like to know how to produce the profiles, please send me a Private Message.)
Calibration result: Accurate scan data.


Negafix

Negafix is the foundation of getting true (ie, accurate, precise, faithful) scans from Negative film.

Negafix's only purpose is to effectively extract the negative film's 'orange mask' from the image data before inversion.

Once the mask has been extracted, the image data should be the equivalent of what we would see if we developed a photograph from the film using conventional projection methods. This means that all colour casts caused by light temperature variables should be retained within the image. Only then would we have a true image.

For the average user, standard Negafix should be sufficiently accurate to produce acceptable images. ('should' would be true if the profiles were accurate which unfortunately is not true for many of the profiles provided with SF Ai NikonM.) For the consummate expert or for those cases where an appropriate film brand/type/speed profile does not exist, Negafix Expert mode allows duplication and customisation of profile mapping.

Negafix result: Accurate image data from negative film.


(Automatic) Color Cast Removal

The Color Cast Removal tool removes colour casts caused by light temperature variables. These might include sunrise, sunset, overcast skies, florescent lighting, tungsten lighting, etc.

Removal of the colour cast is optional. As Ian Lyon points out in his excellent SilverFast tutorials (see the section The Levels Histogram), there are occasions when the colour cast is desirable such as the colour cast caused by a sunset. In these cases, the Color Cast Removal tool would not be used or would only be used to a limited degree (ie, somewhere between 0% and 100%).

Automatic colour cast removal was designed principally as a batch scanning tool. A real-world example of its use: a roll of APS film whose images include:
  • photographs shot outdoors on a sunny day (no cast)
  • photographs shot outdoors on an overcast day (blue-ish cast)
  • photographs shot indoors with florescent lighting ('dirty' green cast)
  • photographs shot indoors with flash lighting (no cast)
Automatic colour cast removal allows SilverFast to automatically remove the different colour casts without intervention by the user. Note that these colour casts are not a characteristic of the film itself, and hence should not be affected in any way by Negafix. Nor should they be affected by calibration.

While the Color Cast Removal tool provides automatic colour cast removal, SilverFast also provides the Global Color Correction tool which allows the user to manually and very efficiently create, edit or remove any colour cast within the image.
    I believe that Color Cast Removal affects the whole dynamic range of the image, not just the end points. Color Cast Removal may not be as absolute in the middle as it is at the end points but it will never the less have a visible effect.

    Color Cast Removal remaps the end points of the RGB colour data which is then redistributed to utilise the full dynamic range of the image subject to the (Auto) Threshold values. This redistribution or stretching of the colour data indirectly affects the mid-tones effectively removing much of the colour cast found there. Pressing the Option key while viewing the histogram data will show that the mid-tones are redistributed along with the end points of the image.
Color Cast Removal result: Accurate colour without the unwanted effects of light temperature variables.


Conclusion

These are my impressions. They may be incorrect. Hopefully, Lasersoft will point out any errors in my logic.

Currently, SilverFast disables calibration when Negafix is used. I (and a few other people in the Tips & Tricks forum) don't agree with this logic but that's the way it is. Hopefully, combining SF Ai and SF HDR will allow us to use both Calibration and Negafix together.


Back to the original problem:

When SilverFast is first installed (very important point), automatic Color Cast Removal will be triggered whenever the Image Auto-adjust tool is used if the Auto: Color Cast Removal value is non-zero, and the CMS:Scanner > Internal value is <NONE>. This is true for both Positive and Negative scan types.

If Scan Type is then set to Positive, CMS: Scanner > Internal is changed to Calibration or ColorSync and a scanner profile is specified via CMS: Scanner Profile, automatic Color Cast Removal will be disabled (the Image Auto-adjust button will be coloured).

At this point, I only perceive a problem in the logic that automatic Color Cast Removal should be disabled when Calibration or ColorSync is enabled. As I have already stated, I believe that Color Cast Removal should be independent of Calibration or ColorSync.

If Scan Type is now reverted to Negative, SilverFast automatically reverts CMS: Scanner > Internal to <NONE> and disables the CMS: Scanner Profile effectively turning Calibration or ColorSync off. Now, even though SilverFast has automatically reverted CMS: Scanner > Internal to <NONE>, automatic Color Cast Removal is still disabled (the Image Auto-adjust button is still coloured).

If we now switch Scan Type back to Positive, set CMS: Scanner > Internal to <NONE> and CMS: Scanner Profile to <NONE>, and then switch Scan Type back to Negative again, we can see that automatic Color Cast Removal has been activated once again.

So what's going on? Does the scanner profile still affect the CMS workflow even when its disabled by the switch to Negative Scan Type? Or is this a bug?


Many thanks to Ian Lyons for his replies. Even though we may have different opinions, his replies stimulate further examination and learning of the SilverFast environment and imaging in general; always a worthy and educational exercise.

P.S. In general, I abhor flashy web technology too, especially HTML email messages and Flash 'enhanced' web sites.




kind regards
Gregory

(time to walk the dogs, take a break and contemplate the meaning of the universe and life itself :wink: )

Re: My impression of the tools

PostPosted: Tue Oct 15, 2002 8:17 pm
by ilyons
Gregory,

Calibration serves to match the colour data received from the scanner with known colour data. It does not perform any colour balancing, mask removal or color cast removal. It ensures that the scan data after digital mapping is as true a representation of the film as possible.

While some people may disagree, I believe that calibration should be independent of Negafix, producing accurate scan data which would include the orange mask if present. Negafix could then be applied to accurate calibrated scan data. If possible, I would prefer to use SF Ai to produce calibrated HDR scans of negative film, and then use SF HDR's Negafix to extract the orange mask and invert the image.


We don't really calibrate the scanner we profile it or measure its colour rendering characteristics. The main difference between negatives and postive type films is the mask. The mask used on negatives can vary GREATLY from film to film - even within the same film family. Add in the complexity of different manufacturers and you should start to appreciate that what we have with negative film is a vast set of possibilities. Okay folk will claim we could have a favoured film and so the equal of an IT8 that represents it. However, the BIG problem with the mask is that it varies with development and to some extent exposure. I've seen situations were the same film type appears totally different depending upon the outfit that processed it. Put simply we are trying to hit a moving target and it simply cannot be done with any degree of certainty about the results. The mask colour isn't a constant and without at least one constant we're beat.

Folk seem not to understand how a scanner sees their negatives or what the scanner app does in the first instance to get rid of the mask. The image is inverted and then the endpoints are set. See the folllowing tutorial for the workflow in Photoshop.

http://www.rgbnet.co.uk/ilyons/tutorial ... al_6_1.htm

Negafix's only purpose is to effectively extract the negative film's 'orange mask' from the image data before inversion.

Once the mask has been extracted, the image data should be the equivalent of what we would see if we developed a photograph from the film using conventional projection methods. This means that all colour casts caused by light temperature variables should be retained within the image. Only then would we have a true image.



No - if you look look a the example I show in the above tutorial you'll see what it looks like after the end points are set. Also note that these end points are UNIQUE to that image (we don't have a constant). The other component of Negafix is the curves and without these the results would be flat!


The Color Cast Removal tool removes colour casts caused by light temperature variables. These might include sunrise, sunset, overcast skies, florescent lighting, tungsten lighting, etc.


ACR deals soley with highlight and shadow casts. Those that reside in the midtones are NOT effected.

Currently, SilverFast disables calibration when Negafix is used. I (and a few other people in the Tips & Tricks forum) don't agree with this logic but that's the way it is. Hopefully, combining SF Ai and SF HDR will allow us to use both Calibration and Negafix together.


Negafix removes the same colour casts as handled by Auto Adjust and ACR combination. You can even use the ACR slider in the same way as the Auto Tolerance in Negafix. Try this:

1. scan a negative as positive

2. open the negative image in photoshop and invert

3. save the positive

4. open the positive in HDR - it should appear flat and have a cast (likley blue/cyan)

5. disable cms in HDR so that you have full access to ACR.

6. look at the three channel Histogram and note the scale or range of the data, close the histogram

7. press the Auto Adjust button and the image should now appear fairly neutral and not nearly as flat as before.

8. have a look at the 3 channel histogram again. Note that the end points have been set tight. Also note that ACR slider is at 100%

9. Drag the ACR slider towards 0 (no correction) see the cast begin to appear!!!!


Now reactivate the IT8 or CMS and repeat step 7. Notice that the colour cast remains as you would expect now open the Histogram and drag the ACR slider from 0 to 100% - notice that you still have colour cast! This cast is caused by the ICC profile making corrections to a base image that no more resembles the IT8 than my backside.


Even after fine tuning with ACR you may have an image that requires a curves adjustment to fix the contrast and midtone colourcast issues - this is what the negafix profiles do.


Stage 1 of NegaFix is effectivley Auto Adjust plus ACR. If you keep ACR active for negatives what can it do that hasn't already been done? THINK !!!!!!!!!!!!!

Please stop looking at dead flies on the windshield and concentrate upon the cars in front - if you don't - Kabooooommmmmmmmm!

beginning to understand...

PostPosted: Wed Oct 16, 2002 6:48 am
by Gregory C
It would appear then that my understand was one of an 'ideal' situation, one which is pretty much impossible because the negative mask is never consistent.

If Negafix is colour balancing each image by matching the end points, why do we need multiple Negafix profiles? Do the curves (hence, contrast, saturation, etc) form a general basis of adjustment for each mask?

Should Auto CCR always be inactive when Negafix is used, irrespective of the CMS status?

The v**s**n manual actually recommends customising its mask removal by scanning a blank frame. If each of my films began with a blank frame; one shot in complete darkness; could I use this to fine tune the Negafix profile? Should I flash photograph a colour target to profile instead? Admittedly, it wouldn't be of professionally controlled colour quality but wouldn't it be a good starting point?

When somebody said that SilverFast features a steep learning curve, they weren't kidding!

regards
Gregory

And the actual bug is...

PostPosted: Wed Oct 16, 2002 6:56 am
by Gregory C
So, the bug that I am seeing is in fact that before any adjustments are made to CMS: Scanner > Internal or CMS: Scanner Profile, CCR is automatically triggered with Image Auto-adjust even though Pos./Neg. is set to Negative and Negafix is active.

Auto Color Cast Removal should by design always be off whenever Negafix is in effect.

Correct?

Re: beginning to understand...

PostPosted: Wed Oct 16, 2002 7:55 am
by ilyons
Gregory,

It would appear then that my understand was one of an 'ideal' situation, one which is pretty much impossible because the negative mask is never consistent.


Thats is pretty much how it is.

If Negafix is colour balancing each image by matching the end points, why do we need multiple Negafix profiles? Do the curves (hence, contrast, saturation, etc) form a general basis of adjustment for each mask?


The endpoints remove the highlight/shadow cast ONLY. Different films will have a characteristic midtone colour cast. Also if you look at the spread of data for each colour channel in the Histogram (NegaFix version) you'll see that the spread of data isn't full-scale or even near it, and that typically the data for each channel is at different parts of the histogram. You would be hard pressed to see such an "ODD" spread of data on any positive type film. Actually if you saw this spread of data on a positive the image would like like crap and you'd be contacting Canon/Nikon for an estimate of costs to fix you camera :)

If you go to the Negafix Curves window you will see that each channel has its own UNIQUE curve. This is the NegafFix profile and it contains the base info to remove the midtone casts and tone curves. It can be fine-tuned to suit your own requirments.

Should Auto CCR always be inactive when Negafix is used, irrespective of the CMS status?


It is inactive. Just because the Auto Adjust icon isn't clear doesn't mean that it is working - it isn't and you can see that from the fact that after pressing Auto Adjust the ACR slider remains at 0% - i.e. No Auto Colour Removal.

The v**s**n manual actually recommends customising its mask removal by scanning a blank frame. If each of my films began with a blank frame; one shot in complete darkness; could I use this to fine tune the Negafix profile? Should I flash photograph a colour target to profile instead? Admittedly, it wouldn't be of professionally controlled colour quality but wouldn't it be a good starting point?


All that v**s**n is doing is providing you a means of FIXING the end point values for base-film. You use these values and theory says a perfect balance - but theory and reality are rarely bed fellows!

The profiles supplied by LaserSoft are also just starting points. You can alter the curves in any way you like and save them out as a new profile name.

When somebody said that SilverFast features a steep learning curve, they weren't kidding!


I agree that SilverFast is complex, but not as complex as some would make out.