Scanner performance - v700 and Vuescan Pro - in relation to step wedges

[swmcl]

Greetings,

I recently scanned some photos trying to get an idea of how dense various parts of the negative really were in an attempt to evaluate my workflow so that I was making negatives that could be either scanned or optically printed one day. I came across a result that made me think about how much influence the software driving the scanner was making. To try to eliminate the scanner software I thought of making a scan with significantly large areas of clear glass and black matting. Those two areas (if significant in size) would be treated by the scanner software as the upper and lower bounds - the two ends of the histogram.

I assume the software will interpret any value in between in a linear fashion. I have turned off all controls on the scanner software and scan to a non-compressed tiff file. Scanning as a 16bit per channel -vs- 8bit per channel has no bearing on the density range. Exactly the same graph will result with a heck of a lot more shades of grey in between.

To relate the scanner results to real life I scanned a calibrated 4x5 Stouffer step wedge that has 31 steps.

Attached to the post are three pdf files. One of a step wedge with areas of clear glass and black matting, one of just the 4x5 step wedge itself and one graph showing the difference between the two.

Six observations / conclusions:

A. The scanner begins to be non-linear at about a density of 1.75 to be generous. This would mean (to me) that one should not develop a negative to a density of greater than 1.75 or so to linearly digitise it. In reality we might choose to go beyond this a little.

B. The scan that has the absolute limits set seems to have a more linear response. It may not matter in reality. Certainly, the histogram in Photoshop has room to move for the slider at the high and also at the low ends. To me a scan that establishes the 'absolutes' of 'white' and 'black' is one that does have much less capability of software interpretation.

C. The film enthusiast should not develop film to have a greater density than about 1.75 if they wish to have a linear scan made to digitise their image.

D. To check the neg density range make sure to scan with large enough clear glass and use a black matt. The digital values -vs- densities are clearly seen in the results I've posted (assuming softwares and scanner hardwares produce a similar result ...)

E. A scanner will scan a bit more density range (linearly) than one could print optically without dodging and burning. Straight print -vs- straight (linear) scan the scanner has more range but only just. A diffuser enlarger density range of 1.2 is a digital range from around 220 (density of ~1.47) to 40 (density of ~0.27) on the chart.

F. Given the fact of (E) one can expose generously and pull development (with a goodly strong developer) knowing that in the event of a slightly denser negative, it is able to be scanned very well. Low density negatives cannot be rescued by either digital scanning nor optical printing.

One question:

How, on the basis of these results, can a scanner manufacturer claim to be able to scan a density of 4.0 ??? The scanner is blocked up at 255 with a density of around 2.4 ! (and it is highly non-linear from 2.0 ...) Perhaps the only thing I can think of is the 'exposure' setting (if there is one). Perhaps letting the CCD collect more light over a longer period of time ? Given my question, it makes no sense whatsoever to buy a higher end scanner based on the 'fact' that it can scan to some fictitious density ...

If others could post their results from their step wedges it would be a great comparison. I'll go out on a limb and bet the results will be very similar across scanners and softwares which means if people include large areas of clear glass and also use a black matt they can interpret their digitised results pretty accurately (no need to buy a Stouffer step wedge after all ...).

And BTW I have contacted Epson about their Epson scan software that is limited in its file size. I have asked them to consider those of us who would scan the whole glass platen without limitations but I don't think they'll change. I think too the Digital ICE might only work on small film sizes in the Epson film holders.

All to the good !

Cheers,

Steve

[Ken Lee]

Here's a scan of a Stouffer Step Wedge on an Epson V700 using Epson Scan software. This test suggests the scanner can distinguish around 16 steps, which means 8 stops of dynamic range. In log terms, this is around 2.4, which should be adequate for b&w film exposed and developed normally.

When I asked Sandy King about this, he pointed out that color transparency film can reach densities over 3.0, so for challenging color work this scanner may not be the best option.

I don't know how linear the result is, but can't that be corrected fairly easily if we apply an appropriate correction curve ?

[swmcl]

Interesting ...

I believe Sandy is correct in that colour slide gets seriously high densities. I have the same sort of figure (3.0) in mind for Velvia.

On the scan above you have no detail at either end ! How did you achieve that ?!

:-)

[Ken Lee]

On the scan above you have no detail at either end ! How did you achieve that ?!

I made that test 4 years ago, but I presume I simply used the Histogram tool in the scanning software, setting the black point to the clear film edge and the white point so that the dense area which makes up the background of the image is pure white, with no further clipping.

The process is described in this article. Using Epson Scan software, it's very easy.

Once you adjust the black and white points, you can control the linearity to some degree by adjusting the Gamma setting. I'm not sure what I used back then. Presumably a Gamma of 1.0 gives a linear scan, but I have not tested this.

The image I posted here is not perfectly linear. If the scanner can discern 16 steps, the resulting image should contain 16 evenly spaced steps reading 0, 15, 31, 48, 63... 255. However, in Photoshop they read 0, 5, 31, 55, 77 etc.

We could make a correction curve in Photoshop to linearize the values.

[vuescan]

Hi,

Just been reading through this. I look after the marketing for VueScan - so cannot say anything at all about this technically (sorry!) but if you did want me to ask Ed Hamrick anything in particular just let me know. Happy to get his opinion on something if that would help??
Regards,
Beverley.

[JaZ99]

I assume the software will interpret any value in between in a linear fashion. I have turned off all controls on the scanner software and scan to a non-compressed tiff file. Scanning as a 16bit per channel -vs- 8bit per channel has no bearing on the density range. Exactly the same graph will result with a heck of a lot more shades of grey in between.

Perhaps you scanned a negative in negative mode. If you want to get linear readings, you have to turn linear mode on (scanning gamma=1.0).

I've published my densities from IT8.7 target here: https://sites.google.com/site/negfix/scan_dens (Epson V750)

JaZ

[swmcl]

Hi JaZ99,

My readings are very linear up to a point I should say, but could you enlighten us on this 'linear mode' ?? Where does one "turn on linear mode" ? I see, 'Image', 'Slide film', 'Color negative', 'B/W negative' and 'Microfilm' and it has to be one of those choices !

I'm all for a process that is fully explained and transparent as possible in that what I want is a WYSIWYG response to my scanning. What I see on film is translated to a digital file without any interpretation.

Ken,

Any curve you make would have to counteract the curve put in there by the scanner response as I understand it. You'd have to essentially generate a polynomial of the scanner response and counteract it. I reckon with electronics being so stable it would be possible for the manufacturer to publish such a polynomial but I can't see it happening. It may even be as simple as a response curve from any other CCD type device - I don't know. All I've ended up doing so far as I know is pin the extremes (black and white) to the extremes of the digital range (0 and 255) and I hope that the scanner software will just linearly sample all the shades in between.

I have done this in practice with a real photo and can say that the scanner software output is dramatically different when the 'absolutes' as I call them are not included in the first high-resolution scan. I do admit that first scan is a huge file but the first action in Photoshop would be to trim it to size leaving all the values intact. I've created files larger than 3.5G for 4x5 so Photoshop CS6 goes to this size at least !

To me, there is no sure footing to the purity argument when you don't know for sure that you are eliminating the assumptions in the software. For me, the whole 'purity' argument falls flat when the process is not known. My art is actually going to be mine and Ed Hamrick's and Epson's art coz they all had a hand in making it and they all acted independently from each other without the other one's knowledge.

Hi Beverley,

I think what would be excellent would be for Ed to enlighten us on how to do an absolutely straight scan that does not interpret anything at all in the image. JaZ99 speaks of a 'linear' mode - so far, I can't find the word 'linear' anywhere in the settings of Vuescan (version 9.5.25 on Linux - professional mode)

Even better would be a simple document enlightening us to the process by which a scan is made so that we can choose our options in an informed way. The document simply need to address the concerns of the LF community which are in the main a desire for high accuracy and purity of result I guess. People here are worried about the process and not just whether the final result is pleasing to the eye.

I would also say there are features that Ed could explain for us like Digital ICE and whether it works (or why it doesn't) and the 'density display' option for starters.

Cheers all.

[JaZ99]

Hi JaZ99,

My readings are very linear up to a point I should say, but could you enlighten us on this 'linear mode' ?? Where does one "turn on linear mode" ? I see, 'Image', 'Slide film', 'Color negative', 'B/W negative' and 'Microfilm' and it has to be one of those choices !

I'm all for a process that is fully explained and transparent as possible in that what I want is a WYSIWYG response to my scanning. What I see on film is translated to a digital file without any interpretation.

Linear mode is the mode when the scanner works in linear fashion, that means, the scanning gamma is 1.0. Since you are using Epson scan you are out of luck, I'm afraid.
Perhaps you can try to linarize the scanned data with new_gamma=1/scanning_gamma.

[djdister]

Linear mode is the mode when the scanner works in linear fashion, that means, the scanning gamma is 1.0. Since you are using Epson scan you are out of luck, I'm afraid.

Not true. Epson Scan allows you to use a straight (linear) gamma or alter it as you wish. It is not however termed "linear mode" so anyone looking for such a button to click will not find one.

[JaZ99]

Not true. Epson Scan allows you to use a straight (linear) gamma or alter it as you wish. It is not however termed "linear mode" so anyone looking for such a button to click will not find one.

Last time I've checked there were two gamma settings only: 1.8 and 2.2. But it was a few yers ago, newer looked back at Epson Scan again. Good to know they update the software.