Scanning and levels

[Charles Twist]

One of the great things about digital cameras, is that they knock out an image which occupies a full range of levels as read by Photoshop: this gives an impression of quality. It might be that the sensor doesn't pick a wide enough palette of colours to be realistic, but at least the infallible computer tells you that it's doing just fine . Now, we all know the virtues and vices of different film stocks when it comes to palettes, but what's bugging me, is that when I scan the sheet of film, even when it has a full range of tones, the levels are not spread across the full width of all possible brightness values, as displayed in the Histogram and as modifiable by Levels adjustments. This means that when projecting the narrow range on to the full width of the histogram, there is a degree of interpolation where the computer inputs the values according to some algorithm. The narrower the initial range, the more questionable the end results (and on negative film, that can be very narrow). To what extent is an adjustment of Levels a reliable method? Is there a better way? Is there a way of selecting the range of brightness levels to be scanned at the scanning stage so as to make the process more reliable? For information, I use Epson's own scanning software to produce raw files. There might be cleverer software out there...
Thank you for your help.
Regards,
Charles

[numnutz]

I have wondered about this... I think that the scanner, (mine is an Epson 4990) reads from the brightest part of the scan area and the darkest to set itself up for the scan. The Epson 4990 claims its DMax is around 4 (from memory). I didn't like the software supplied with it so I use Vuescan. I think this gives me better results than the supplied Epson software but It seems to take two or three scans sometimes to get a reasonable result.

So I wonder if the scanner sets up the histogram as being between zero no illumination and white full illumination with the negative being positioned somewhere in the middle of the Histogram. The Vuescan software allows scanning at up to 48 bit I am unsure as to any difference this makes when reducing to 16 or 8 bit levels.

BTW I scan in Mono I haven't tried colour as yet.


nn

[gary mulder]

I do not know if it will help you but most drumscanner software actually controls the analog digital conversion. Resulting in a nice histogram. The software from imacon appears to do something similar. But if it is possible to really control a CCD that way I am not sure. And maybe it is a post digital conversion.

[scovell001]

hello Charles,

I'm trying to respond to your post without sounding condescending, so apologies if I do.

Its very easy to get tonal values (o-255) confused with colour gamut. The histogram shows the distribution of tones within an image on a scale where 0= pure black & 255= pure white (I'm sure you already know this). It doesn't show the gamut (colour volume) that the device (camera/scanner) can capture. When you see the histogram on a digital camera display, its showing you how the CCD has recorded certain values. For example, you can see a brighter white than a digicam cdd. But the ccd can only cover a certain dynamic range. The ccd records what it thinks is a pure white, (because thats as far as it can see into the highlights) and maps that to 255 (and 0 for black) on the camera histogram. So, you think you've got this a nice full histogram, but in reality full dynamic range has been lost, ie some whites have been clipped & blacks are plugged. In reality, you'd want to have a narrower readout histogram, to ensure a digicam has captured the full dynamics of the scene, similar to that which you are experiencing with your scanning. You can then go and set highlight & shadow (0 & 255) yourself later.

Back to scanning.
In photoshop (and all other software) 0 is pure black, & 255 is pure white (as you already know). But, your Epson scanner can't record a true 0 black natively (I don't there is a scanner that can). So, what your scanner thinks is pure black, is actually around 15-20 (in the case of an Epson) on the 0-255 scale. Your scanner does have excellent highlight dynamics, so again what is a pure white in the negtaive/tran would normally record around the 230-245 mark (unless you've inadvertently clipped the highlights in the scanning software). Therefore, depending on the dynamic range (daytime, nightime etc) of the original neg/tran, this will determine where its tonality sits on the histogram & give you that narrow range in the middle of a 0-255 histogram scale aka - a small lump in the middle of a huge black desert!

The final part would be to stretch that narrow band of information in the middle of the histogram so it fills the full histogram bar. You can do that simply by setting highlight & shadow in levels, in Photoshop. Hold down the alt key and drag each of the highlight and shadow sliders in until they meet the edges of the histogram but don't clip any colours. Hit OK, and now look at the histogram bar. You'll see that the histogram is now full from 0-255.

So, to summarise. There is no way to set which brightness levels your scanner scans. It will just record what it thinks is pure black and pure white (and all in between) to a histogram which illustrates black as 0 and white as 255. Its up to you to adjust the highlight & shadow to map the image to the full width of the histogram display.

Hope this helps

Ian

[Charles Twist]

Hello Ian,
I am glad you took the time to answer, given your experience. While I knew many of the points you make, I didn't realise that scanners were tied to an inherent dynamic range. In spectroscopy, I am used to PM tubes where you set the voltage to control the dynamic range! What your answer says is that scanners will not do a good job of colour negative film, since there are no blacks or whites, only shades of russet. Very disappointing.
Thanks for your help.
Best regards,
Charles

[Patrick Dixon]

I'm not an expert on scanners but I know a bit about digital TV, and I'm not convinced that's correct. The sensors have inherent characteristics that you can't change, but if you set the 'black' (lowest) and 'white' (highest) points correctly in the scan software it should make the best use of the digital dynamic range - which for 16-bit values (48-bit colour) would be 96dB. If you allow the scanner to do it own thing and the image's dynamic range only occupies a small proportion of the digital range, then applying a levels adjustment later in PS will introduce granularity through the interpolation process. You see this in the PS histogram which becomes a series of spikes rather than a complete range.

And 'black' and 'white' are concepts as far as imaging goes - it's up to you to define what they mean depending on what you're looking for.

[Joanna Carter]

I have always found adjusting the "black" and "white" levels in the scanner software introduces less "grain" than doing it in Photoshop.

[scovell001]

Hopefully these 2 documents will help provide the definitive answers to all those with queries about scanners/scanning:
http://www.hutchcolor.com/PDF/Scanning_Guide.pdf
http://www.hutchcolor.com/PDF/Cneg.pdf

[Marizu]

In Joanna's Photoshop course, we covered selecting the appropriate dynamic range for the particular film and its exposure on the V750. Doing this did seem to produce images that had a better 'shape' in Photoshop than doing the expansion after the scan.
We weren't sure whether the Epson software was just doing the expansion of whether something different was happening in the scanner. What we were sure of is that different noises came from the scanner if we set different ranges for the same neg so it 'sounded' as if it was doing something different.

[Patrick Dixon]

CCDs are analogue (amplitude) devices, so I'd expect the software to be setting gain and level in the stage prior to the analogue to digital conversion. I'm not entirely sure how that would make the scanner 'sound' different though.

[dave_whatever]

I thought I'd bump this topic since I've just aquired a 4990.

After just a couple of test scans (16/48bit) it seems to evidence what other people have said, the setting the levels in epson scan gives you smoother shadows than doing the same in photoshop to an uncorrected scan. Now of course the puzzling factor is that for a scanner with allegedly no analogue gain, the two methods should give identical results.

Anyone know what is going on here? Either the scanner does have gain, or the epson software is doing something sly when outputting the file, a bit of compression maybe?

[scovell001]

Perhaps someone would be kind enough to post for download or, send me the following:

A transparency scan they made in Epson Scan
A scan of the same transparency in Silverfast
The raw HDR scan (scanned in Silverfast) along with their ICC profile for their scanner.

I'll have a look at whats going on with Chromix and come back with the definitive answer.

I'm using the Imacon & I've made drum scans using the levels process as mentioned in this forum. I have no noise in my scans. There must be something happening in the Epson software maybe.

[dave_whatever]

Unfortunately I don't have a working version of SF, or means for scanner profiling yet, hopefully someone else will pick this up Ian. Thanks.

I must say the noise in the shadows is subtle, but noticeable at 100%. It may not even be noticeable in prints, but its always good to know what it really going on.

[Joanna Carter]

Hi Dave

I have the means for profiling screen, scanner and printer. If you don't mind paying my petrol costs, etc, I would be prepared to come and set things up for you. Unless you would be interested in attending one of my "Photoshop for LF Photographers" workshops, where you will not only learn to scan and adjust images but, also, get your scanner profiled there.

[dave_whatever]

Thanks for the generous offer Joanna but if you knew the number of bugs my one year old is bringing home at the moment you wouldn't volunteering to come anywhere near our house! I do intend to get profiled up to my eyeballs in the fullness of time though (I'm already good for monitor profiling, and have no printer), so should I get stuck I'll let you know.

I only got the scanner on monday, spare room still covered with bubble wrap and packaging, I was really just intrigued as to why two processes which should give identical results actually give different results. A bit of a "that's all very well in practice but how does that work in theory?" moment.