Setting levels

[Barry Kirsten]

Hi all. Can someone give me some guidance please about setting levels, particularly input levels? I've been following Ken Lee's excellent notes on scanning, but am curious about a couple of things.

Firstly, in setting the output levels, we choose 0-255. I assume this is to preserve a full range of tones in the output, correct?

Now in setting the input levels, we are similarly seeking to capture all available tones from the negative. Herein lies my problem. In the histogram below I found I had to set the input levels at 19 and 222, to have a manageable print. Setting the upper level at 'b' produced an image which was almost blacked out. So what's the significance of this far end of the curve, and why does including it have such a profound effect on the image?

Similarly, at the other end of the curve the same may be true. In this example adjusting the lower level to 'c' seems to reduce some of the high tones, although the effect is not as strong as that at the other end of the curve.

The tone curve viewer seems to indicate the amount of linearity in tonal structure throughout the range, and I've noticed that the shape of the curve can be modified by moving the centre arrow. What is the common wisdom relating to making changes to the histogram before scanning?

Any help very welcome.

Baz.

[mdm]

I would set the black and the white sliders so they do not clip any data, and then use the middle slider to change the mid tones until the tones in the image look OK. Then you can do some fine adjustment later in photoshop or equivalent. Dont forget about the mid tones.

Sometimes you can clip the extremes if your neg is not masked fully or if the film holder borders are visible.

[Greg Miller]

What you say you are observing does not really make sense.

The right slider controls your highlights. Moving this slider to point "a" should cause clipping (to 255) of all highlight tones that appear in the histogram to the right of point "a". It should not really affect your shadows nearly as much as your highlights.

The left slider controls your shadows. Moving this slider to point "c" should cause clipping (to 0) of all shadow tones that appear in the histogram to the left of point "c".

The middle slider controls midtones. Moving it left (like you have to a value of 1.4) should cause the midtones to be brighter. The overall appearance of the image will typically look brighter. 1.4 is a pretty significant change and would typically only be used for a severely underexposed image.

The main goal in scanning is to preserve all detail, while also allowing for the retention all tonal gradations. If this were my image, I would set the right slider so that it nears, but does not touch (to avoid any accidental clipping), the right most values that exist in the histogram (in this image that would mean moving the slider to just to the right of point "b"). I would set the left slider to a value slightly left of where you currently have it. I would set the middle slider so that a midtone gray appears midtone.

[jim kitchen]

Dear Barry,

I have a few set rules that I apply when scanning with the Epson software, and the rules are as follows:

1. I never use an odd final digit in any of the boxes;

2. I set the gamma (middle pointer) to the gamma of my calibrated screen to gain consistency throughout all of my scanned negatives;

3. I set the scanner's curve to linear only, again to achieve scanning consistency, using the second icon from the top, then I close that window once it is set;

4. I activate the densitometer to see what whether any limits are breached while I cruise through the adjusted scan, browsing over any highlights and, or low level areas. The densitometer will tell you quickly if you breach the output limits, by turning the numeric values red;

5. I will adjust the lower input level and higher input level to maximize the scanned information while viewing the adjusted results through the densitometer tool.

6. I never set the Output level to more than 250, ever…

Practice and practice some more, but you will find that the densitometer tool is very effective at magnifying any area of concern, and you will recognize that minor adjustments to the input levels will keep your scanned negative within bounds.

Just my two pennies… :)

Have fun,


jim k

[Greg Miller]

6. I never set the Output level to more than 250, ever…


I'm curious as to why you would choose to limit highlight detail at the scanning stage. It is those highest order bytes that can hold the most information. Why not carry that detail into Photoshop and then adjust the white point at time of output based on the characteristics of the output medium?

[jim kitchen]

Dear Greg,

I knew you would ask me that... :)

It is an old school habit and a habit that I perpetuate in PS too, because I notice that the following does happen:

1. during a scan the highlights can and do get washed out when one sets the output limit to the maximum value, and I would rather have detail in specific highlight areas compared to not to use as fit for purpose. Setting the limit to 250 and monitoring those limits using the densitometer tool confirms my assumption that the highlight details can and do disappear, although I never tweak the specular highlights, so I try to avoid that issue with careful monitoring of the altered scan. I now have a scanned file that I can make brighter if I desire to do so in PS, and I know that the highlight information is present in the digital file, save the specular highlights;

2. during any exercise in PS the biggest mistake that many users tend to make happen is their continuous effort to maximize the upper output values to 255, thinking that they are accomplishing a full range of values across the finished print. They do, but with what I consider poor guidance. Again, I use the same logic as scanning, because the final image produces a file with zero information in the areas where the PS value is better than 250, which can lead to gloss differential across the paper base. The printer does not lay down any ink onto the paper base in those zones of 250 or greater. I would rather have PS lay down some thin ink layer in the highlight areas, albeit a weak layer of ink, compared to not to reduce gloss differential, and again, if the highlights are not just right in the finished print, I tweak the highlights by selectively making their PS value climb closer to 250 and, or slightly above, where that is a rare event, and never anywhere near 255. Sometimes I go as low as 240 to achieve the desired effect when the highlight situation dictates that I must do that in the finished print.

My way of transferring old school "flashing the paper" crap taken into PS to control the highlights effectively.

jim k

[Greg Miller]

Thanks Jim. I appreciate the explanation.

[Barry Kirsten]

Many thanks David, Greg and Jim for your assistance. There is so much to learn about digital processing, but I'm determined to get on top of it (no alternative since I don't have a darkroom any more!). Every little bit of info helps, and what you've said is logical and very helpful. Thank you taking the time to help.

Barry.

[MumbleyJoe]

Jim - I have a couple follow-ups if you'd kindly indulgent me. (I have now owned my own scanner for roughly 2 weeks, so I am by no means an expert, but I spend more time fretting about details than I should)

1 - I'm really trying to understand why you limit the input value on the highlights, as Greg was asking about. The way I'm thinking about it is that if your highlights appear to safe to cut off at, let's say, 230 and you leave it open to 255 - sure, you're going to have no detail above 230 to work with (and your densitometer readings would show that). I guess the reason why I have been leaning towards scanning "wide open" (to 255) is that those densitomer readings and histogram are based on the quick preview scan. I don't know exactly how accurate that preview is, but I like to err on the side of assuming that there may be a little more detail on the fringes of the histogram than it captured in the preview. If you cut the input off too low, it's possible to clip those details in the final scan. My preference is to scan the full input range and then open it up in Photoshop with a levels adjustment (well, I use curves to achieve it, but "levels" is a more obvious idea) watching for what areas are actually going to clip first. (In this case of course clipping could be at the maximum 255, or at a lower output value as you suggest, which I totally understand).

Anyway... I'm just trying to make sense of your approach over what I'm doing. I'm just at the beginning of my own scanning phase, so I'm just trying to balance the best practices of others with my own. (I'm a fan of your work too, so it's not as though I wish to challenge your methods, merely understand them).

2 - Never entering an odd final digit. Really? Off hand I can't really think why that would be, but feel like there's probably a perfectly rational reason for it, so I'm dying to know.

3 - Finally, the gamma. You set it to your screen setting (2.2?). I typically just leave this at 1 (for now), but then, gamma seems to be the one property that constantly really baffles me. The output at anything drastically different from 1 seems to me to be substantially "off" from what the final image requires (which is fine if that's the best method to capture detail, etc).


/sorry to threadjack

[jim kitchen]

Dear Tyler,

Have no fear, anybody can challenge my methods, because scanning is a black art at best, and my approach is visual while using the extremely limited densitometer window, along with some experience how the Epson scanning software works, whether one considers it to be a preview scan or not. I will describe my logic as best as I can, and why I want the results that I obtain. My process should not be considered to be better than any other process, but just an alternative approach… :)

Please note that I realize we are limited to any visual output, with the tools presented to us through Epson, but experience directs my scanning process too. You should also note that my negatives are developed to accommodate my scanning process and my expected results, so your mileage and your logic may be pressed a wee bit to understand why I do the things that I do while scanning, but I will try express the process to the questions you asked.

To begin, I want a scanned file that mimics the tonal range of my negative that I want to see not what the scanner software says I should have, where the information is contained within a histogram that is somewhat balanced, and somewhat tonally flat, compared to any histogram skewed to the left or right and, or clipped across the top. I want a digital file that I will adjust later in PS, compared to a few folks mindset that one must get everything perfect in the original scan, and if you don't you are hooped. I do find that mindset to be somewhat limited in their approach, and I just view it as an alternative process, compared to what I employ. As I said earlier it is "Voodoo," just like some economic models today. I post-process in PS to produce an image that I want, compared to an image that I am forced to live with or make a straight print from, because the scanner approaches and sets the negative's original limits differently. Everyone has their own style of scanning, and although each method is probably good, better, or best, my scanning method just happens to be a visual one through the data eye of the densitometer. As a side note, there is one feature that I have always wanted to design for Epson as a third-party plugin, and that plugin would be a new window that shows the histogram in larger more significant detail, complete with instantaneous feedback in a separate preview window upon the original scan, as one changes the input, output, and gamma values. My desire would be to create a tool that adds value and fines tunes the black art visually before you press the scan button. That will happen someday, by someone smarter than I, and quicker to market that I.

So, back to your input and output question…

You, along with many other scanning souls, myself included, do not have the tools to see instant gratification within the finished result, so we guess while sliding the end points, the gamma setting, and the output levels about, hoping that we set them properly. A wing and a prayer sometimes… My only salvation is to review the densitometer reading upon a set displayed data points within its window to see whether my modifications to the slider controls made a significant change and, or a minor change to the displayed pixels. Sometimes we are too technical in our approach to obtain perfect initial results, so much so, that we forget we have a secondary option in PS to correct any minor errors that occur during a scan. We never had this luxury in the darkroom except for flashing the paper, and that is where my logic originates. So, if you never flashed paper in the darkroom to control the highlights in an over-processed negative, then you might fail to completely understand why I even use or employ this scanning process, but then again you might not. Using Barry's thumbnail for instance, you will note that he discussed the "a" and "b" setting issues while trying to obtain a decent scan, and I would like to talk to that issue. I have a set routine, designed to accommodate my negatives which are processed for a scanner, so that I can minimize my factorial responses when I tweak any single numeric value in the window. I set my known output boundary limits prior to tweaking any input limits, including the "Tone Correction" curve, where I neutralize any affect it may have upon the scanning process by setting the "Tone Curve Name" to "Linear" and close the window. If I were to follow along with Barry's example, not knowing the effect that other controls have upon the output I would be stymied, because I would not know which setting was best and, or correct. Barry realized that his upper input limits had a great effect upon the tonality of the final scan, as he should, but not knowing where to set the final points properly may produce a scan that happens to be a get-by-product at best. Having some background knowledge would short circuit that get-by-product event quickly.

So to clarify my baseline settings they are as follows:

Output: 0 and 250
Upper and lower Inputs: as the curve dictates, and from experience...
Gamma: 2.20 because of my negative development process...
Tone Curve Name: Linear

I should note that I do set the lower limit to 10 and not zero (0) periodically to accommodate a scan that seems to benefit visually from the increased setting. Old school scanning would have told you to never go below 10 as the lowest setting because oldschool told you that papers could not produce a darker value below 10, so why scan a value below a value of 10?

Knowing that I only must play with the upper and lower limits of the histogram to determine the information I retrieve from the scan eases my pain. When I adjust the lower end of the histogram, such as the dark retrieval area, I review the data in the densitometer window to see whether I made the major dark tones too light or not light enough. The same is true at the other end of the histogram, where I review the modified scan continuously to see whether my upper input highlight setting breached my output limit of 250 or not, and whether I have wiggle room to make the highlights a wee bit lighter. That limits my exploration, once the other boundaries are set.

Going back to Barry's upper input limits issue, if he had preset his gamma value, his upper and lower output limits, and his tone curve name, Barry would have seen in a heartbeat that lowering the setting to "222" would brighten his image compared to not, and he would only need to verify that the "222" setting was the value he desired, compared to say "224" or "220" while using the densitometer value window. Using this approach should satisfy your angst about clipping the values in the original scan. If I clipped a few points on either end of the scan's histogram, but it produced a scan that I absolutely adore, who cares, especially if your clipped areas are lost in the shadows and the paper white when you print. My goal is to keep the highlight values within a region that prints well compared to not, save the specular highlights. The deepest black areas will fall where they may, but controlling the lower input level helps you with your lower mid tone dark values, as visualized in the densitometer window too. Keeping your highlight values below the paper white will allow you to produce an image that is not de-popularized by gloss differential, especially if you continue that mindset into PS and do not let your upper output values exceed 250.

As for the even digit versus odd digit… That is old school PS dictating what I do in any application, since a bug did exist and may still exist within PS that was caused by odd digits, and since I program computers that control oil and gas processes, it is just another mindset that I have when division occurs in any application while using an odd numeric value. Funny as it may seem, I just avoid odd digits in any calculation, especially if I did not write the software.

Lastly, your question about gamma… This is a value that I set because it conforms to my desired brightness output level, caused by my negative's development process, and not so much as my monitor's gamma. It seems that they just coincide nicely, so I tend to discuss it as monitor gamma. Gamma is an important part of the scanning process, and unfortunately a part that seems to be misunderstood by many. The consistent nature of my developed "scanner" negatives dictates that I use a gamma value of 2.2. I never vary from that, but I will probably do so once I change the film, my developer, my scanner and, or any other inline variable that might interfere with my original empirically discovered process settings.

Again, these are just my observations from my own negative developing and scanning process, where I am certain that you will obtain values that are completely different because your development process is different, your equipment is different, and your printing style is different. The information should allow you to tighten your scanning process, and hopefully minimize your experimentation procedures. Your continued experience with the scanning process, your equipment, and your printing style will dictate your results going forward.

If you have further questions, just ask… :)

jim k

[Ken Lee]

You can include a little portion of the film edge, and adjust the left-hand slider to clip it at Zone 0.

However, do not always place the right-hand slider at the right end of the curve. Not all subjects contain pure white values. Below is a recent example.

Perhaps my statements have been unclear. I follow the same approach as cinematographers: start with the subject itself. When perfectly lit, exposure is effortless, and so are developing and printing: resulting images have a natural analog quality. If we can't control the subject, then the next best thing is exposure and development. If we can't get ideal exposure and development, we can correct it while scanning or in the darkroom. If we can't get the scan right, then we can correct it in Photoshop. The ideal is to make corrections as far upstream as possible.

Ansel summarized this approach far more concisely: "When asked what camera I use, I reply 'The heaviest one I can carry'."


http://www.kenleegallery.com/images/forum/img401t.jpg
Massachusetts 2010
Kodak 2D, 450mm Fujinon C
5x7 HP5+, Pyrocat HD

[Greg Miller]

1 - I'm really trying to understand why you limit the input value on the highlights, as Greg was asking about. The way I'm thinking about it is that if your highlights appear to safe to cut off at, let's say, 230 and you leave it open to 255 - sure, you're going to have no detail above 230 to work with (and your densitometer readings would show that). I guess the reason why I have been leaning towards scanning "wide open" (to 255) is that those densitomer readings and histogram are based on the quick preview scan. I don't know exactly how accurate that preview is, but I like to err on the side of assuming that there may be a little more detail on the fringes of the histogram than it captured in the preview. If you cut the input off too low, it's possible to clip those details in the final scan. My preference is to scan the full input range and then open it up in Photoshop with a levels adjustment (well, I use curves to achieve it, but "levels" is a more obvious idea) watching for what areas are actually going to clip first. (In this case of course clipping could be at the maximum 255, or at a lower output value as you suggest, which I totally understand).

Note that Jim is not setting the "input" value for highlights to 250, he is setting the "output" value for highlights to 250. So he is not clipping highlights. An input value of 255 would be mapped to 250 (and all values below 255 would be remapped as well).

There is no right or wrong here. My preference is to do this step in Photoshop when preparing an image for output. That is because different output media (monitor, matte paper, glossy paper) have different characteristics and capabilities, and those characteristics will change (usually for the better) over time. I prefer to leave the output value at 255 so that I can keep brighter whites for preparing an image for my website and set a different value when printing on different papers, and also maintain the maximum amount of information for if/when the output media improve for the better.

Of course this only matters for images that have information greater than 250.

I also change the output for blacks during printing (not during scanning). I have printed a step wedge for each of my paper/ink combinations to find the lowest value of black that I can print and still hold detail. For example, with my Epson 4800 and Hahnemuhle German Etching paper, any value below 9 will render as the same value of black (and therefore will not have detail between 0 and 9). So during printing I set the black output value at 9 if (and only if) the image contains black values below 9.

[bob carnie]

I like Jims explanation, everyone has a different method for sure.

One little tip that I like to do is activate the digital Colour Meter . app in all my Macs
I set the eyedropper to minimum and use the lab setting.. If you are more comfortable with RGB numbers then set it there, or if you are more comfortable with CMYK settings then set it there.

To Jims point , I will set the density in the high and low end to points that I know will create tone in significant detail areas that I want to print.
I must say that usually I move the sliders out side the clipping areas to leave as much detail but I use the meter to make sure I have the density readings for paper I am going to print.

This meter is used all the time to verify colour and I will wave it over any and all areas of images , just to practice my eye to see and understand colours/density using the digital meter. I find it to be a wonderful little tool to use to help look for colour inbalances when scanning negs as well.

I do not know if this app. is available for PC.
I wish someone could adapt this device to read original scenes.

[jim kitchen]

Folks,

I would like to take a brief moment to clarify a comment I made earlier about tweaking the scanner's upper input value to make sure that it does not breach the output value that I set within a scan. I will use Barry's example again… :)

I mentioned that I peg the scanner's output value to 250 for reasons stated earlier, but I also mentioned in passing that I adjust the upper input value to verify that the scanner's upper output value does not surpass 250, while using the densitometer tool. For example, I locate a few pixels or a larger pixel grouping that I know I want to control in the finished scan, and where I do not want any of the pixels in that grouping to migrate past 250. I locate that grouping by scouring the pre-scan within areas that are logically bright and look bright, such as a cloud formation, a hand and, or a bright forehead in a portrait.

Upon locating that grouping, I discover that if I left my setting at "222" and, or at position "a" as illustrated within Barry's example, I might see that the illustrated values in the densitometer window are red in colour when they are greater than 250. This is a clear signal that my current upper input limit at "222" is a wee bit too robust, and I must adjust my upper input limit to a higher value, which happens to be in the direction of position "b," therefore making the red density values disappear, and therefore making the entire scanned image darker by two, three, or four density points. I do believe that the densitometer software is intelligent enough to warn you that your upper input limit setting requires further adjustment because a few pixels exist that will be clipped. I do notice that the clipped values are present in the scanned file, when I do not adjust the upper input limit properly. This upper input limit tweak would surely be at the discretion of the author regarding whether the adjustment is required or not and, or whether the adjustment adds value to their scanned image. Sometimes we become too technical, but sometimes being too technical adds value.

Anyway, I thought I should include this minor fact to allow you to explore the software a bit more thoroughly…


jim k

[MumbleyJoe]

Thanks for all of the clarifying points (Jim and others) - I think I'm more on the same page with you than it may have initially appeared. It is, indeed, a bit of a black art and certainly don't think there's a single approach to fit everyone's needs - if there was, I would kindly ask that someone tell me those precise settings :)

To date, my general approach has been to limit my output to the ~10-245 range, but generally keep the input range at its fullest. This may not give me the best looking scan, but gives me a good working file to take into Photoshop where I'm much more comfortable adjusting the levels to where I think they ultimately belong. I find those adjustments in Photoshop trivial, but have not been nearly as comfortable trying to set those values in the scanning software itself (though I have not really played with the densitometer yet) - so I am conservative with my scanner settings.

Thanks again - I'll probably read through everything yet again along with the example and see if anything still trips me up. The bottom line is producing a result that works for you... which reminds me, I should go post something in the landscape images thread. :)

Cheers!

[MumbleyJoe]

However, do not always place the right-hand slider at the right end of the curve. Not all subjects contain pure white values.

Indeed! This is partly why I tent to open up the input slider to its highest setting, as the initial points selected by the software tend to try to fill the output range.

[williamtheis]

OK advice but IMHO it doesn't go near far enough... far better to NOTHING in the scan & get a RAW file. Most scanners are 12 bit so a 16 bit RAW is very dark... highest values are no more than 50 (note: not 250!)... do NOTHING in the scanner or RAW converter and bring the image into Photoshop where the control is much better

[mdm]

I must admit, I like the results I have got from adjusting a raw scan, but it is more work so I usually just accept what vuescan tells me.

David


OK advice but IMHO it doesn't go near far enough... far better to NOTHING in the scan & get a RAW file. Most scanners are 12 bit so a 16 bit RAW is very dark... highest values are no more than 50 (note: not 250!)... do NOTHING in the scanner or RAW converter and bring the image into Photoshop where the control is much better

[Bob McCarthy]

OK advice but IMHO it doesn't go near far enough... far better to NOTHING in the scan & get a RAW file. Most scanners are 12 bit so a 16 bit RAW is very dark... highest values are no more than 50 (note: not 250!)... do NOTHING in the scanner or RAW converter and bring the image into Photoshop where the control is much better

The scale you quote ie. 0 to 256 represents the scale of black to white in 8 bit, higher bit scales are overlaid and the actual numbers are not used but assumed.

Bob