The "Perceptual" approche - a new Digital Negative method

As Niranjan suggested in another thread I am opening a new thread to discuss a different, new approach in making Digital Negatives.

I've tried many "calibration" methods for the alternative processes and none of them gave good results in any situation. Most of the time you have to re-trace the curves in PS and make more tests to get a good result. I ended-up with the conclusion that the process of "calibration" itself must be, somehow, wrong or incomplete. Calibration tries to map an extended scale of equidistant grays into one with the same number of grays that keeps the equidistance while respecting the Dmax and the Dmin capabilities of the alternative process/paper involved. Simply saying the "calibration" is a method to compress a gray-scale into another gray-scale keeping the relationships between the grays. At first look it seams a correct approche and mathematically it is, undoubtedly, correct.

But...
As many have noticed it doesn't always lead to good results. Why? IMO, the "Calibration" method is a compressing method very similar with the DOLBY sistem in music (the compress of an extended scale of frequencies to meet the limited recording capabilities of the magnetic tape) with an essential difference - when you play the tape the DOLBY system expands back the scale of frequencies and the sound is perceived correctly. This final stage is completely absent in the Digital Negative "Calibration" methods. It is like playing a DOLBY-recorded tape on a non-DOLBY device - the result is somehow flat, un-natural. That's why the existing "Calibration" methods lead to better results if you start with an image with higher contrast.
The mathematical correctness and the process automation can't automatically lead to a perceptual correct result. In music, on a non-DOLBY device a tape recorded without DOLBY sounds better than one recorded with DOLBY. That's because the frequencies are specially adjusted in studio to meet the tape capabilities - not by trying to record them all (impossible on tape) but by restricting to the ones that ca be recorded on the specific tape and by changing the relation between them in order to make the music sound natural. In Architecture the proportions and straight lines of the Parthenon have had to be altered in order to look correct. If they were kept correct they wouldn't look correct. It's a general principle in Arts - things must LOOK RIGHT and not necessary BE RIGHT, contrary to science/engineering. The same should be true in making a Digital Negative for the Alternative Processes. IMO, a completely different, perceptual approach is necessary.

In the last months I have worked at developing the work-flow of such an "perceptual" approach in the creation of the Digital Negatives. It's now in the extended testing stage and, at least for me, it works better and it's more predictable than the traditional ones.

What do you think on the matter?

You might just have a point but how much will you charge the end user?...

I have tried several different methods of crafting digital negatives for pt/pd printing, but have had the most success with this:

Dead Link Removed

FWIW, I have tried a perceptual approach based on a supposedly simplified system that I bought long ago; endlessly tweaking the curve in PS all to no avail. Personally, I like a bit of calibration because it tailors the negative to the print process. That said, I wish you all the best with your system.

Dan, thanks for starting the thread..

As I said earlier, I too noticed this flattening of the image soon as I made my first successful POP print, to my great disappointment, even though I thought I had done a good job of figuring out the correction curve as well the other parameters of the digital negative. Not being much of an audiophile, I didn't make the analogy with Dolby. Unlike Dolby though, we do not have the luxury of expanding back the dynamic range after the transformation with a correction curve is made - simply because we are limited physically by those Dmax and Dmin beyond which we can not go.

It is easy to see on Photoshop what would happen if you take an image and apply the ideal characteristic curve, taking into account the Dmax and Dmin of the process. To counter this, what I did was to apply a contrast curve to the original image, my perceptual approach if you will, before the whole process of correction curve, flip, invert and colorize. And the results were much better, felt in perception nearly as good as the original image. In essence, my approach is not an alternate one but one that involves an additional simple step. It does require that one come up with a good correction curve first with whatever method one wishes to use.

I will break here and continue on another post later to elaborate.

This sounds really strange to me, but I am interesting what you find.
In my world of thinking I want EVERYTHING to be standardised and the only variability is what I do in photoshop. I have ( personal ) found that when I have everything calibrated, scanning equipment, screen, ( the light level and colour temp in the room ), printer and chemical process then I can with high accuracy do changes in photoshop that follow over to the final print and the print that I get in the end match the image that I have on my screen.

Also, your DD example is a bit flawed, DD uses lossy audio codecs! So when you decode the music it wont "expands back the scale of frequencies and the sound is perceived correctly", what is lost is lost and cant be reproduced, now its very smart to throw away the bits that you have a hard part hearing, but whats not there cant be reproduced ( well, you have to beat the Nyquist theorem and so for no-one has.)

When you have something more concrete please write up a post on how to follow your workflow, still very interesting what you find.

Dan Pavel said:

Calibration tries to map an extended scale of equidistant grays into one with the same number of grays that keeps the equidistance while respecting the Dmax and the Dmin capabilities of the alternative process/paper involved. Simply saying the "calibration" is a method to compress a gray-scale into another gray-scale keeping the relationships between the grays. At first look it seams a correct approach and mathematically it is, undoubtedly, correct.

Click to expand...

Does it really? I thought it takes not an extended range but the actual range you have from Steps 1 to 21, and adjust it so that the steps in between Steps 1 and 21 are evenly spaced. If such is the case, there is no compression and your Dolby noise reduction metaphor is misplaced.

I think perceptual calibration is valid. When I get too wrapped up in getting the "perfect" digital neg my trying to all the values to linearize, I don't see the forest from the trees. Sometimes my densitometer is my worst enemy. I'll get a cyanotype that is well printed, but I'll see is as not good because of a value is not "in it's place". If you think of JPEG compression, the image is compressed according to how the image looks. I'm always trying to perfect my craft of alternative process printing. But while trying to achieve perfection in my prints, I end up more of a nit picky scientist. I want the science and the craft to serve my art. This is my 2¢ worth,

Mainecoonmaniac said:

But while trying to achieve perfection in my prints, I end up more of a nit picky scientist. I want the science and the craft to serve my art. This is my 2¢ worth,

Click to expand...

Well then, let it do so.

ced said:

You might just have a point but how much will you charge the end user?...

Click to expand...

My intention is to share it for free and I hope that people here with more knowledge and experience than me will help at improving it. Just give me some time to finish the job and to find a form to present it in a comprehensible and simple manner (taking into account the fact that English is not my native language).

sandholm said:

Also, your DD example is a bit flawed, DD uses lossy audio codecs! So when you decode the music it wont "expands back the scale of frequencies and the sound is perceived correctly", what is lost is lost and cant be reproduced, now its very smart to throw away the bits that you have a hard part hearing, but whats not there cant be reproduced ( well, you have to beat the Nyquist theorem and so for no-one has.)

Click to expand...

I didn't say reproduced correctly, I said perceived correctly. What's lost in the process is not signifiant enough to make the output perceived incorrectly, otherwise the DOLBY system would be pointless.
Even something is lost in the two processes, the principle of the two processes remains the same even their mechanisms difer. Sorry, I can't see the logical flow.

The aim of calibrating a digital negative on inkjet or silver for that matter is to give you a fighting chance of creating repeatable good negatives... In Silver my Lambda calibrates the silver film and lasers to a good linearized 21 step before I print. - the process will just not work .. This allows me to look at L values and easily know that a 94 value will end up printing as a highlight with detail in my final print and so on down the 100 L steps.
That way I can place density predicabley onto film as my file or image dictates - a understanding of how the PROCESS needs in contrast and density is the second more subjective factor.
The role of quad-tone rip or any program out there to set up inkjet digital negatives is to place a predictable density on that film, and to date I have not seen a program designed by anyone that is user friendly, My good fortune was to have Ron Reeder visit us and set up our system here years ago..

I would be careful with this idea of perceptually placing contrast till at least you can build up a proper profile and I do know its not easy and is very frustrating for many workers.

Richard Boutwell by the way has a good concept and is trying to produce a step by step program that would work for dummies like myself

All I can say is go for it!

faberryman said:

Does it really? I thought it takes not an extended range but the actual range you have from Steps 1 to 21, and adjust it so that the steps in between Steps 1 and 21 are evenly spaced. If such is the case, there is no compression and your Dolby noise reduction metaphor is misplaced.

Click to expand...

The number of steps is not relevant. In DN calibration the grays scale (any number of steps) between Dmin-Dmax is compressed to match the different (smaller) Dmin-Dmax interval of the process/paper involved. The initial range is "extended" compared with the range of the final output (what's recorded on tape in the case of DOLBY) in both DN and DOLBY compression processes.

If that is the case, then I see the calibration process as like the zone system where you tailor the negative to fit the paper through N- and N+ development (scientific vs perceptual). I'd be interested in learning more about the "perceptual" approach though. Whose perception? I control contrast with my PtPd prints both in how I tweak the digital negative on screen in LR/PS and with Na2. The calibration process where you linearize the transitions from dark to light is designed to help translate what you see on the screen (linear) to what you see in the print. How do you perceptually calibrate your monitor so that what you seen on-screen matches your perceptually modified DN and thus print.

The work-flow I propose consists, in principal, in displaying on the monitor screen an image as close to the final print as possible. Instead of applying a curve to your "perfect" image, print the negative and develop the print to see the result you'll have on the monitor an image of your final print where you can see what to expect in the end. And that image on the screen is in the real colors of your final print/paper process (very useful for most "colored" alternative processes). You can make any adjustment (use the whole power of the PS arsenal) to your initial image and instantly see the result to be expected in the final print on your monitor. Of course, your monitor (and scaner, if you use one) need to be well calibrated. The "Perceptual" approche is increasingly useful with those processes producing a tinted output or/and an output with a more restricted Dmin-Dmax range.
You can even start with a curve produced with ChartThrob or other automation tool and instantly see on the screen what to expect in the final print. You can modify it or add other adjustments and see on the screen how the'll modify your final print.

You still have to apply a curve to translate your initial "perfect" image in PS into a image of the final print in PS, but it's a different curve than the classical ones (it's somehow their inverse). When you actually print the DN you'll have to disable that curve.
It is actually not that complicated. I am now working on a way to automate the creation of the curve via PS Scripts and at finalizing the step-by-step work-flow in PS.

IMO, the advantages of this method are:
- WYSIWYG
- More predictable results, at least in the more "colored" processes.
- You ca use the whole PS arsenal while instantly previewing on the monitor the results in the final print.
- The testing stage, if not completely eliminated, is much reduced, leading to an economy in paper/chemicals/Pictorico.

Perhaps it is a language barrier, but you still haven't said what the perceptual approach to creating a curve actually consist of.

I've checked in the dictionary. Perceptual = of, based on, or involving perception. Perception = the ability to see, hear, or become aware of something through the senses.
The perceptual approach is not in creating the curve. The curve is created as usually, based on printing a grey scale, etc.... But the curve is not used to create the DN. It needs to be disabled, or you could even delete it if you want, before you start printing the DN.
The DN adjustments needed in order to match your printing intentions are made while you instant visually check their effects on the final print on the screen. That's perceptual. I couldn't find a better name for this approche.

Perceptual means eye-balling it?

Yes, you could say that.

Continued from Post #4:

To illustrate the concept, let's take a digital image in its final form shown opened in Photoshop on Figure 1. We wish to do a contact print with utilizing an alternative process of choice. Let's say after doing the due diligence in the process figuring out the best processing conditions, the darkest block reads de-saturated RGB of (30-30-30) in the scanner and the brightest one reads the same as the bare paper, which is (235-235-235). This establishes the Dmax and the Dmin of the process. Next we do the complete process for making the correction curve with whatever methodology that is convenient. In the end, let's assume you have a perfect correction curve. What I mean by perfect is when you repeat the process after incorporating this curve in the digital negative, the follow-up correction curve is more or less (nothing is perfect!) a straight line from (0,0) to (255,255). The result of using this "perfect" correction curve on the selected image is shown on Figure 2. In Photoshop it can be simulated by making a new layer and apply a straight-line curve starting at (0,Dmax=30) and end at (255,Dmin=235). As can be easily seen, and as would be expected, the resulting image is a flat, dull representation of the original image.

How do we improve on this? We have to add some contrast to the image without losing the details in the shadows or blowing out the highlights. I would say a normal tendency is to think that the correction curve should be fine-tuned a little bit, a new digital negative made and image printed again. That can lead to several time/material-consuming frustrating round trips. In addition, the chances are once it works on an image, the same modified curve may not work for another that has a different character to it.

A better way is to use what I call a pre-curve to simulate the desired outcome before making a negative. This can be achieved by adding a Curves layer above the image background layer and below the characteristic curve layer of the Figure 2. Different shapes of the curves can be tested by checking the resulting image. Make sure that the characteristic curve layer is also turned on at the same time. I have found that by pulling the shadows down somewhat (making sure they do not flatten out) without changing the highlights much gives enough perception of the contrast that was in the original image. Highlights can be very difficult to deal with particularly in an image with clouds, so it is bets to leave them alone. Aesthetically, blown highlights are far more distracting than some loss of details in the shadows.

Figure 3 is the result of adding such a layer with the pre-curve applied. If you cycle from Figure 1 to 2 to 3 and back to 1, you will notice that the image in Figure 3 is perceptively very similar to the original image in Figure 1 even though their histograms are quite different. One curve has worked for most of my images so far but it can be easily personalized for every image.

EDIT: Added the label "Characteristic Curve" in Figure 2.

Continued to another post....

I am not seeing how that is any different than just doing your normal adjustments, whether by curve or slider, in LR/PS and applying your correction curve when printing the DN. The correction curve doesn't defeat your normal adjustments, it implements them based on how your alternative process responds. My resulting images are not "flat, dull representation of the original image". What method are you using to create your correction curve?

I think you fundamentally misunderstand what a correction curve is. If your correction curve is a straight line as show in Figure 2, it is not a correction curve, and there is no need for it. A correction curve is a curve which translates the adjustments curve on screen to the same adjustment curve for the print, taking into account how your process responds.

Continued form Post #20:

Once the pre-curve is defined for the process and saved in Photoshop, the procedure for making the digital negative is as follows:

1. Finish your digital image as you like.
2. Make a new Curves layer and apply the pre-curve.
3. make another Curves layer above. Apply the correction curve.
4. Flip, invert to make negative image.
5. Colorize (if necessary)
6. Print the digital negative

Figure 4 is the scanned image of the final print obtained with this process made on Centennial POP toned with selenium. Compare with the original image – different tones but, I believe, closely matching the perception of the dynamic range. You can find more of my POP images at the link below.

:Niranjan.

Edit: Corrected some mistakes.

nmp said:

Continued form Post #20:

Once the characteristic curve and the pre-curve are defined for the process and saved in Photoshop, the procedure for making the digital negative is as follows:

1. Finish your digital image as you like.
2. Make a new Curves layer and apply the characteristic curve.
3. make another Curves layer below. Apply the pre-curve.
4. Flip, invert to make negative image.
5. Colorize (if necessary)
6. Print the digital negative

Figure 4 is the scanned image of the final print obtained with this process made on Centennial POP toned with selenium. Compare with the original image – different tones but, I believe, closely matching the perception of the dynamic range. You can find more of my POP images at the link below.

:Niranjan.

Click to expand...

hi niranjan
centenial POP hasn't been available for over a decade ..
im a bit confused ... are you editing a scan so it looks like contact printed image
or editing a scan and inverting it, to create a film ( or paper ) negative to be contact printed
on an alternative process ( i use alternative to include contact printing on silver papers as well ).
sorry for my confusion ...

OP
i haven't really delved into the digital negative theory as much as you and others have
when i make them, i eyeball my image on a the computer screen ...
i realize that cyanotypes ( that is usually what i make them for ) don't like a flat negative
so i invert the image and boost the contrast a little bit and bring a cd or memory stick to my local
xerox/copy shop who prints it out for me. if i had a laser printer i would just print it out here
and not bother with the copy shop ... i don't use a densitometer or even think about words like dmin or max
if the image doesn't hav enough contrast i just roll with it, if it has too much, i do the same ...

Because it's true that "A picture is worth a thousand words" I did a little test.
An Alternative Process/paper combination can be simulated in PS by a curve. I simulated VDB on Fabriano Artistico with a curve saved as "VDB-FabArt.acv" (I already had it om my HD). It simulates a print made with a DN with no correction applied,
Applying this curve to an image will simulate a VDB print on the Fabriano Artistico paper. The simulated print is an ideal print, with no variable induced by the scaner, chemicals, temperature, etc.
Let's take an image and apply the curve to it. The result will simulate a VDB print with no correction applied to the DN.

Now let's use a Calibration tool (I've used ChartThrob) and make it produce a curve for this VDB ideal process.
1-ask ChartThrob to create its step_wedge
2-apply the "VDB-FabArt.acv" curve to the ChartThrob step_wedge.
3-ask ChartThrob to produce the curve. Save the curve as "ChartThrob-VDB.acv".

ChartThrob Calibration
1-open in PS the initial B/W image
2-apply the "ChartThrob-VDB.acv" curve to it.
3-apply the "VDB-FabArt.acv" to the result.
This will simulate a VDB print with a DN calibrated with ChartThrob.

The Perceptual Approach
1-open in PS the initial B/W image
2-add a curves adjustments layer
3-load the "VDB-FabArt.acv" to it. It needs to always remain the top-most layer, no matter how many other layers you'll add!
4-come back to the initial image layer and use as many adjustments layers, filters, etc you want to make what you see on the screen look the way you want the final print to look like, but always keep the "VDB-FabArt.acv" curves layer the top-most and enabled.
Here I've used a modified ChartThrob curve, a "Silver Effex Pro2" filter and a "Brightness/Contrast" adjustments layer.
This will simulate a VDB print with a DN made with the Perceptual Approach.

Here are the results showing the difference between the two approches:


On images with greater contrast ChartThrob is doing a little better, but still I can get better results with a Perceptual Approach.


You can do the test on your own images and with your own calibration tool. You can download the "VDB-FabArt.acv" here: https://drive.google.com/open?id=0B4dENtC-VAIheGtBS0FmRlRSTFk

faberryman said:

If your correction curve is a straight line as show in Figure 2, it is not a correction curve. A correction curve is a curve which translates to adjustments curve on screen to the same adjustment curve for the print, taking into account how your process to responds.

Click to expand...

It is not the correction curve. I am not showing the correction curve for the process anywhere. What I am showing on Figure 2 is the "characteristic curve" as I call it. Correction curve corrects the negative to linearize the map of which the characteristic curve is the result. For a theoretically perfect correction curve, the characteristic curve is a straight line between Dmax and Dmin.

I added a label in Figure 2 to reflect this.