Full Histogram Normalization in scans: it is always right?

[eriksatie]

I'm scanning a lot of old photos and slides with Silverfast and an Epson V800. I'm done with the negatives : ) they was about 1500, it takes my about 1 year. The scanner is profiled with IT8 targets bought from LSI using the auto profiling function inside Silverfast.

With negatives it was not needed to normalize the histogram because Negafix do it anyway, so after applying it you have a scan that is more or less normalized. but with prints and slides you always have a big parte of the histogram (in lower and in the higher part) that is basically unused. Sometimes it doesn't look so bad in the monitor, sometimes it so strong that you need to normalize it, at least a bit. I'm talking about normalization of the 3 RGB channel together, not tweaking with each channel differently.

But a full normalization (without clipping) it's not always nice to see: shadows can get too strong, or highlights get too bright, and contrast can really go high. Also colors change a little bit. Measuring LAB values I've noticed that normalization is not color neutral. L channel changes (obviously) but also A and B channels change.

So what's the best approach?

1) fully normalize the histogram and then correct for contrast and color shift. In this way you are using the full spectrum of the histogram for having better color gradients
3) normalize just for what looks good and eventually let some portion of the histogram empty in the left or in the right (adding contrast could have a similar result)

[saiguy]

Try the approach I suggested in your post under Color Management. This here is the better place to post as the scanning geeks may weigh in here.

Hit ACC then go to the Histogram tool. Most scans will pull the right sliders to where you would normally think to place them. And that will remove casts in the highlights and set a good white point, especially if there are nice white clouds for SF8 to key on. If you do this on all scans you will find the right side highlight positions way to the right on some images. Somehow SF8 is super smart at this. If you see the highlight sliders way to the right, and you move them to beginning of the HL data, thats where you will find the blow out and over exposure you are experiencing. I trust SF setting of the white point 99.9% of the time. If there are specular HL it can trick SF and then I just increase the top slider in the Picture Settings tool and move on.

You need to set the black side sliders in the Histogram tool per channel. Not with the Master Tab. If SF moves them at all with ACC, it will move all 3 to where ever the data starts on whichever channel is most to the left.

Then use Selective CC to reduce over saturated colors, can shift the Hue a bit also if needed.

You should go for a good base scan. They will all require post processing in an app. such as LR if you want optimized photos. Don't try to set contrast in SF. Setting the white and black points will give you a file that you can do contrast tweaks easily in post scan processing.

[digitaldog]

I don't understand the use of: fully normalize the histogram.
Unlike in a raw converter where you've got to 'normalize' the image shot optimally for raw data (because the current rendering assumes something different), in the scanning software, the Histogram should represent what it really is. You want good color and tone appearance and you can't always find that produces a 'nice looking' Histogram. Use the tools in the scanning interface to provide a color and tone appearance you desire globally.

[eriksatie]

I don't understand the use of: fully normalize the histogram.
Unlike in a raw converter where you've got to 'normalize' the image shot optimally for raw data (because the current rendering assumes something different), in the scanning software, the Histogram should represent what it really is. You want good color and tone appearance and you can't always find that produces a 'nice looking' Histogram. Use the tools in the scanning interface to provide a color and tone appearance you desire globally.

I'm posting an example of this. In attachment there is the histogram of a photo print. If you look at the print it is ok, no underexposed, dark shadow (one subject has a black t-shirt). If you look at the scan it is not totally bad, but one could say that it is a bit dull, it lacks contrast, blacks are not blacks, highlights could be pushed more. The histogram clearly show this. Almost every scan I made from prints have this characteristics.

Personally I'm not sure why this happens, but I think this is due to the characteristic of the photographic paper. Paper is not perfectly white and cannot produce a totally white point. and black is not really black.

So what is the best approach to treat it? can I just move the left histogram slider to the right just near the beginning of the histogram? and the same for the left slider (but to the left)? is this a "good practice" that always works well? There shouldn't be any loss of data, but results are not always pleasing

[eriksatie]

Try the approach I suggested in your post under Color Management. This here is the better place to post as the scanning geeks may weigh in here.

Hit ACC then go to the Histogram tool. Most scans will pull the right sliders to where you would normally think to place them. And that will remove casts in the highlights and set a good white point, especially if there are nice white clouds for SF8 to key on. If you do this on all scans you will find the right side highlight positions way to the right on some images. Somehow SF8 is super smart at this. If you see the highlight sliders way to the right, and you move them to beginning of the HL data, thats where you will find the blow out and over exposure you are experiencing. I trust SF setting of the white point 99.9% of the time. If there are specular HL it can trick SF and then I just increase the top slider in the Picture Settings tool and move on.

You need to set the black side sliders in the Histogram tool per channel. Not with the Master Tab. If SF moves them at all with ACC, it will move all 3 to where ever the data starts on whichever channel is most to the left.

Then use Selective CC to reduce over saturated colors, can shift the Hue a bit also if needed.

You should go for a good base scan. They will all require post processing in an app. such as LR if you want optimized photos. Don't try to set contrast in SF. Setting the white and black points will give you a file that you can do contrast tweaks easily in post scan processing.

I'm starting using ACC in SF8. Since now I've almost scanned negatives and it wasn't needed, because Negafix takes care of neutralizing the orange mask, and doing that, also the histogram was usually ok. I'm seeing that it works quite well, but I am interested in understand what's the best "manual" approach, to learn how to deal with this kind of image. I have no problem using automatic functions but I want to fully understand how they works, because sometimes they doesn't work the way you want

Why you said that you *need* to set the black point per channel? doing this I will change the color balance and that's not always the case. Why is it wrong to do it in the master channel?

[saiguy]

Why you said that you *need* to set the black point per channel? doing this I will change the color balance and that's not always the case. Why is it wrong to do it in the master channel?

Eriksatie,

ACC will set a good neutral white point. It will not set a good nor neutral black point. That is why you need to use the Separate Tab in Histogram to do this per channel. It will fix, neutralize, the color balance of the darkest zones. The mid tones is a different story. I use Neutral Pipet and Selective CC to deal with that as best I can do so with SF. Using the Master Tab is not best practice.

You mentioned before that the Neutral Pipet works if you can find a grey place to click. It can work almost anywhere. But you need to set in Pref's Auto that the tool will work in that way and not turn the clicked place to RGB128. You posted a photo with yellow clouds. Using the NP would probably have made the clouds white. But it works on a somewhat narrow range, with the clouds lets say from 240 to 255. It will not remove yellow cast globally. But it can be used again in areas with lower numbers. If you had used ACC the yellow clouds would have been fixed most probably.

SF8 is complex and hard to learn. If you want to take the deep dive get Make Seagal's eBook available from LSI. It is very searchable.

I have done about 15,000 scans, mostly 35mm mounted slides, and about 300 paper photos. Have gone thru Mark's book fully at least twice, and  boned up on some sections several times. You talk about paper types. I think it doesn't matter, just get the best base scan you can get.

Take the photo with the yellow clouds and follow my instructions and see what you get. They ALL still need further processing in a post editing app if you want good results. SF8 will not deliver a finished photo. I think you are looking for some simple few clicks solution for your scans. It will never happen like that.

If you are not willing to endure the learning curve or do not have the time to do so, Epson has a scanner, fast photo or something like that, that does super fast scans on paper photos. May produce better results than what you are getting. It is about $600. PhotoshopCafe.com has a very good review of this product.

[digitaldog]

I'm posting an example of this. In attachment there is the histogram of a photo print. If you look at the print it is ok, no underexposed, dark shadow (one subject has a black t-shirt). If you look at the scan it is not totally bad, but one could say that it is a bit dull, it lacks contrast, blacks are not blacks, highlights could be pushed more. The histogram clearly show this. Almost every scan I made from prints have this characteristics.

Well the image clearly shows it as you state, so don't use those settings; fix the lack of contrast and push the blacks.

[eriksatie]

Why you said that you *need* to set the black point per channel? doing this I will change the color balance and that's not always the case. Why is it wrong to do it in the master channel?

Eriksatie,

ACC will set a good neutral white point. It will not set a good nor neutral black point. That is why you need to use the Separate Tab in Histogram to do this per channel. It will fix, neutralize, the color balance of the darkest zones. The mid tones is a different story. I use Neutral Pipet and Selective CC to deal with that as best I can do so with SF. Using the Master Tab is not best practice.

You mentioned before that the Neutral Pipet works if you can find a grey place to click. It can work almost anywhere. But you need to set in Pref's Auto that the tool will work in that way and not turn the clicked place to RGB128. You posted a photo with yellow clouds. Using the NP would probably have made the clouds white. But it works on a somewhat narrow range, with the clouds lets say from 240 to 255. It will not remove yellow cast globally. But it can be used again in areas with lower numbers. If you had used ACC the yellow clouds would have been fixed most probably.

SF8 is complex and hard to learn. If you want to take the deep dive get Make Seagal's eBook available from LSI. It is very searchable.

I have done about 15,000 scans, mostly 35mm mounted slides, and about 300 paper photos. Have gone thru Mark's book fully at least twice, and  boned up on some sections several times. You talk about paper types. I think it doesn't matter, just get the best base scan you can get.

Take the photo with the yellow clouds and follow my instructions and see what you get. They ALL still need further processing in a post editing app if you want good results. SF8 will not deliver a finished photo. I think you are looking for some simple few clicks solution for your scans. It will never happen like that.

If you are not willing to endure the learning curve or do not have the time to do so, Epson has a scanner, fast photo or something like that, that does super fast scans on paper photos. May produce better results than what you are getting. It is about $600. PhotoshopCafe.com has a very good review of this product.

Thanks for all the details. Probably you skip it, but as I wrote in my first post (the other thread) I’ve bought and read the Mark’s book and find it very helpful to fully understand SF8.

really I’m not looking for a one click solution. I scanned about 1.500 35mm negatives and I’m happy with the results. now I’m starting with positive and prints, and I hoped that these media would be easier to deal with. No orange mask, no inversion, no film profile...

I’m just looking for an optimized workflow. I’m trying to figure out if I can speed up certain processes, maybe create some automation with Affinity Photo, what I need to do in SF8, what can be done outside it in a faster way. There’s a filter in Affinty Photo that can normalize the histogram and I can run it in batch on all the scan, I’m just trying to understand if it is a good idea

[JaapD]

The histogram of the scan looks good to me: no clipped highlights and shadows which is the main goad for a proper scan. You did scan in 16 bit per color, right? Now it needs to be post processed, in Photoshop or Affinity Photo. Here you can shift the blacks to the left in order to create proper blacks to your liking. Same approach with the highlights. Then you’ll probably need to apply an S-curve to make the picture less flat.


Regards,
Jaap.

[Guillermo Luijk]

The histogram is not the goal. A good looking image is.

If you normalize here the histogram to spread to both ends, you'd ruin the atmosphere (just a test image, not meaning it is good at all):

.

Regards

[digitaldog]

The histogram is not the goal. A good looking image is.

+1!

[digitaldog]

And.....

Everything you thought you wanted to know about Histograms

Another exhaustive 40 minute video examining:
What are histograms. In Photoshop, ACR, Lightroom.
Histograms: clipping color and tones, color spaces and color gamut.
Histogram and Photoshop’s Level’s command.
Histograms don’t tell us our images are good (examples).
Misconceptions about histograms. How they lie.
Histograms and Expose To The Right (ETTR).
Are histograms useful and if so, how?

Low rez (YouTube): http://www.youtube.com/watch?v=EjPsP4HhHhE
High rez: http://digitaldog.net/files/Histogram_Video.mov

[eriksatie]

Very interesting video and answered to lots of my questions

can I ask for some more : ) ?

1) during the video you said that LR does a better job "expanding" the histogram of an 8 bit image. How can do it? does it apply some sort of tonal interpolation?

2) you talk a about histogram manipulation on luminosity or RGB histogram, what about single channel expansion/compression? why and when use it?

And.....

Everything you thought you wanted to know about Histograms

Another exhaustive 40 minute video examining:
What are histograms. In Photoshop, ACR, Lightroom.
Histograms: clipping color and tones, color spaces and color gamut.
Histogram and Photoshop’s Level’s command.
Histograms don’t tell us our images are good (examples).
Misconceptions about histograms. How they lie.
Histograms and Expose To The Right (ETTR).
Are histograms useful and if so, how?

Low rez (YouTube): http://www.youtube.com/watch?v=EjPsP4HhHhE
High rez: http://digitaldog.net/files/Histogram_Video.mov

[digitaldog]

Very interesting video and answered to lots of my questions

can I ask for some more : ) ?

1) during the video you said that LR does a better job "expanding" the histogram of an 8 bit image. How can do it? does it apply some sort of tonal interpolation?

2) you talk a about histogram manipulation on luminosity or RGB histogram, what about single channel expansion/compression? why and when use it?

1. I did? Where in the video are you referring. All processing of 8-bit per color data takes place in high bit within LR/ACR's processing engine. Maybe that's what you're referring to?
2. If you edit a single channel, based on a Histogram or not, you're going to alter the color. I don't think you want to be doing that.