Vuescan vs Epson Scan -software

[Dave Humphries]

Hi Manfred,

Thanks for your replies to my queries.

the scans and restorations are going to potentially happen over a period of years. He is also looking at using two different scanners for the process.
~
I have noticed slight colour differences in the output of my two scanners (same brand, different model).

You're right, I had not thought that far ahead

I this kind of work, there can be some fairly heavy duty adjustments, to bring back the colours by working on a channel by channel basis. These adjustments are definitely more extreme than what one would encounter in normal day-to-day PP work from a digital camera capture (unless the exposure was totally blown). I found I had to spend a fair bit of time cleaning up artifacts, that I suspect might have been the result of using an 8-bit image. I expect that the issues were likely due to pushing the PP software to extremes and building on additive errors of the various incremental steps I was taking.

I suspect that an 8-bit jpeg will be more than sufficient for 95% of the images I've worked on. The problem is that other 5%, and frankly I can't tell by just looking at the image if problems are going to arise in the restoration work.

In respect of bit depth for scans, to clarify, if I were doing it, the very first task upon opening the 8 bits/channel original scan would be to convert it to a 16 bit workspace so that I don't lose any data when post processing - and (of course), save it with a different file name - rule #1; NEVER overwrite an original file.

AFAIK, almost any manipulation of say, exposure values, is going to result in the resultant math on given pixels exceeding 8 bit depth. If we just do a few operations, it probably won't be noticeable 95% of the time, but as you also say, you cannot visually predict which ones will be 'compromised' by reduced bit depth, so safer to work 16 bit/channel, as indeed was my intention.

With storage being relatively cheap, I would tend to scan to TIFF and stick with 16-bit sRGB data.

True enough, and it may be simpler (and safer if likely to forget to up-convert bit depth), to start as you mean to go on.

I just seemed to be in a "why do we do that?" mood last night

I ran into a situation early last year where someone asked me to redo a colour wedding photo that was both faded and stained and all they had was someones jpeg scan for me to work with (no original available). If I recall correctly, the scan was using sRGB.

I this kind of work, there can be some fairly heavy duty adjustments, to bring back the colours by working on a channel by channel basis. These adjustments are definitely more extreme than what one would encounter in normal day-to-day PP work from a digital camera capture (unless the exposure was totally blown). I found I had to spend a fair bit of time cleaning up artifacts, that I suspect might have been the result of using an 8-bit image. I expect that the issues were likely due to pushing the PP software to extremes and building on additive errors of the various incremental steps I was taking.

The artefacts were probably because of the capture jpg coding process (and compounded by any subsequent jpg processing that might have been done) rather than bit depth, especially if they were 'edge related' - IOW the change between adjacent image content that is dark and light in tone. That said; if the artefacts were posterisation, then yes, it was a bit depth issue. Chances are, both will apply though!

As long as the image is being edited in 16 bit depth and remains open in the image editor, errors of bit depth and jpg artefacts should not accumulate, even if we repeatedly Save (as jpg). The artefacts should only become additive if the image is repeatedly cycled through "saved > closed > opened > edited > saved" operations. The most damage is caused by the conversion to 8 bit and the jpg coding process, but this only occurs (with best practice) to the exported file, not what is in the image editor's memory. I made this point recently in another thread here (by a different member) on photographing their family archive of documents and photos.

A related analogy:
I guess many of us have photocopied documents in our lifetime - if we need 10 copies, we put the original on the glass and photocopy it 10 times, we don't do one, use that as the original for the next copy, and so on, because we know that the tenth copy will be 'dire' - but if we repeatedly cycled through "jpg saved > file closed > jpg opened > edited > jpg saved" operations, that's what we're doing.


Thanks, Dave

[stighenning]

Thanx Manfred, Dave for your details in explaining this issue. Hopefully others will find this thread/information useful too.
I have continued my search for better images..
Came across a page that wrote about taking "photo of the picture"..
So I mounted my tripod, camera + lens + Canon 500D Close-up Lens and a really bad lighting (I know :-)). But its for the test..




This is what I done:
Left image is the "original image from the scanner". Right image is from the Canon 7D Mark II.


Closer up:
Left image is the "original image from the scanner". Right image is from the Canon 7D Mark II.


And with Faststone Viewer 5.5 - and "auto adjust color". Left image has 49+ colors and the right image has 130+ colors.
again.. Left image is the "original image from the scanner". Right image is from the Canon 7D Mark II.

[stighenning]

Photo taken on the kitchen floor..
I know not the right place, but Its testing purpose.

JPG


RAW file and adding color

[Manfred M]

Your bottom two images show the problems with your technique quite clearly. The images are not flat and the light distribution is quite uneven,

[jnielsen]

Interesting post. I like Vuescan. However I am very impressed with the Epson "Color correction" example you posted. I do not use Vuescans "Restore Color" option. I find it makes very strange colors, exactly like in your example.

I normally do not use Vuescan "RAW". Only sometimes for trying different settings in Vuescan without actually scanning the picure again. Then you can "scan" it from the RAW file.

I do not Vuescan for processing the pictures. It means I use Vuescan to scan in 48 BIT "neutral" TIF. This "neutral" TIF is not the same as the RAW file above. It looks slightly different. The main settings in Vuescan are
Filter: Infrared clean = light (default options)
Color: Color balance = None
Output: TIFF file enabled, TIFF file type = 48 bit
Then in a second step I batch process in Lightroom or AcDsee Pro. Mainly Auto levels and sharpness. And when needed white balance and color correction. And then I saves as 24 bit JPG.
Here are some of my scanning examples ( slides ) made this way. However not from slides with faded colors.
http://www.datanord.dk/dias-scanning...dias-scanning/

This normally gives good results. However I found red faded pictures like yours very difficult to correct. After working with curves and sliders I normally end up with a simple manual white balance and exposure/contrast.

I tried to load your Epson scan picture and the picture taken with the camera and process with AcDsee and Lightroom. And it seems that some of the better result of the Epsons picture is because that the original scan is better (more color information to work with) like if the Epson program catches more colors out of the picture than the Vuescan program and the Camera. Maybe some other settings in Vuescan would help.

[stighenning]

Hi everybody! (or for those of you that read this )
I've been testing a bounch of software.

Epson V700 with Epson scanner software
Epson V700 with vuescan
EPson V700 with Silverfast AI demo suite
Epson V700 with third party software

Canon 9000f Mark II with canon software
Canon 9000f Mark II with vuescan software
Canon 9000f Mark II with Silverfast AI demo suite
Canon 9000f Mark II with thirdparty software using twain or WIA drivers.

Third party scanning software:
Autosplitter 1.7.6 (actually testing 1.7.7 beta - and this working out REALLY GOOD!)
Scanspeed - this seems to hang..

Conclusion:
So I startet using Autosplitter with WIA Epson driver. And the photos looks really good! Scanning 600dpi, 24bit, tiff-lzw, and the photos are as good as Epson Scanner software - and much faster to use.
I took 3 different photos. Scanned with both printers, different software. Compared them and zoomed into the picture.
At first - Canon scan looks better with more "colors", but after looking.. you see that there "is to much color".
I though Silverfast would match up, but it did'nt. Remember that I'm scanning photos (and not negatives).

***
Autoscanner:
I just scanned 207 photos in 2h 13min. 1 Album scanned
(using a batch method, and putting the images in same place on scanner)


***
Another topic...

Also selecting which application to handle/view all 200k photos/videos:
Faststone Image Viewer (using this at the moment)
IrFanView 64 4.44
XnViewMP
XnView
digiKam
Zoner Photo Studio X
StudioLine Photo Basic 4
Phototecha
Photo Supreme

[Shadowman]

Hi everybody! (or for those of you that read this )
I've been testing a bounch of software.

Epson V700 with Epson scanner software
Epson V700 with vuescan
EPson V700 with Silverfast AI demo suite
Epson V700 with third party software

Canon 9000f Mark II with canon software
Canon 9000f Mark II with vuescan software
Canon 9000f Mark II with Silverfast AI demo suite
Canon 9000f Mark II with thirdparty software using twain or WIA drivers.

Third party scanning software:
Autosplitter 1.7.6 (actually testing 1.7.7 beta - and this working out REALLY GOOD!)
Scanspeed - this seems to hang..

Conclusion:
So I startet using Autosplitter with WIA Epson driver. And the photos looks really good! Scanning 600dpi, 24bit, tiff-lzw, and the photos are as good as Epson Scanner software - and much faster to use.
I took 3 different photos. Scanned with both printers, different software. Compared them and zoomed into the picture.
At first - Canon scan looks better with more "colors", but after looking.. you see that there "is to much color".
I though Silverfast would match up, but it did'nt. Remember that I'm scanning photos (and not negatives).

***
Autoscanner:
I just scanned 207 photos in 2h 13min. 1 Album scanned
(using a batch method, and putting the images in same place on scanner)


***
Another topic...

Also selecting which application to handle/view all 200k photos/videos:
Faststone Image Viewer (using this at the moment)
IrFanView 64 4.44
XnViewMP
XnView
digiKam
Zoner Photo Studio X
StudioLine Photo Basic 4
Phototecha
Photo Supreme

Nice follow up to the original post. 14 months or so later still using the same scanner which is a rarity these days as new OS can kill old gear. Haven't done many scans where quality was an issue, just been doing some scans to archive some news articles/photos. If I get around to shooting some film I'll look more deeply into scanning unless I get the film digitized at the processor.

[stighenning]

More research.

I read that scanning in 48bit first, and then converting into 24bit later on would/can create better photos, than scanning directly with 24bit.. I have check this on several photos, and in my opinion, scanning directly into 24bit is faster AND also greats better images. 48bit scan most of the times creates photos with more artifacts than 24bit does.

***

After testing with Epson Scan software again.. I see that Epson DOES create better images WHEN SCANNING ALL PHOTOS ONE AT THE TIME (or several photos on the glas, but the scanner has to scan each photo separat, going back and forth)

I now see that AutoSplitter does scan with "no color" even though the autosplit images does look good.

What I did;
When I use Epson Scan with "color control".


Then I add all photos into scanner,
click preview,
and I have to manually select each photo. (time consuming.. the autofindphoto does not always work)
Click scan.


Positive
The photos get an even better color, looks better AND have much more colors to work with..

Negative
- The big minus (-) is that the scanner have to "go back en forth" for each photo it is scanning, and takes much more time AND guessing it will wear-out the scanner faster :-(
I even tried having the scan page up - leaving the old selection (where the photos where placed) and put in new photos on the scanner, and clicked scan directly... BUT this does not make any good scans... The only way is to select "preview" and do the mark-select-pictures all over again to get the right color adjustment for each photo. :-(
CORRECTION: I found the Save settings.. Doh! Just enter a new name, and click save -and it will save the selections too. (The settings include the exact position of the marquees)

[stighenning]

MORE testing..

Turns out after investigating further, that the photos are still best scanned "one at a time" with Epson Scan.
Much more colors into the picture. The picture is also more "richer" and looks better.

I did purchase both Autosplitter and ScanSpeeder.

I've testet ScanSpeeder some more. And this actually creates better images than Autosplitter. And by that I mean it creates images with more colors. (doing color count on the image, after they have been scanned)


So my workflow is as follows:

1.) for important pictures, use epson scan software and scan each photo separatly, 600dpi and 24bit.

2.) for bulk loading images; place 3 photos on the scanner, with 1-2 cm between them.
Then I use epson scan software to scan the whole scanner plate into "on large tiff file".
Open up ScanSplitter and makes it do it magic to cut out the images into files.

[Shadowman]

Epson software does seem to have a lot of bells and whistles that actually do something, I was testing scanning some magazine images and what gets pulled the TIFF looks nothing like what's on the page (that's not to say the mags quality is top notch). but the editing software really takes it to something visually usable.

[stighenning]

Epson software does seem to have a lot of bells and whistles that actually do something, I was testing scanning some magazine images and what gets pulled the TIFF looks nothing like what's on the page (that's not to say the mags quality is top notch). but the editing software really takes it to something visually usable.

See if it makes a difference when scanning with epson software, if you select "within the image" and not outside the image your scanning. (do not select to much outside the white scanning area)

[Round Tuit]

Hi again Manfred,

Another thing that's bothering me about discussing such things as bit depths is that if these are paper prints being scanned, isn't the dynamic range of the images produced limited to around 4-5 stops? i.e. the reflective range of light on paper

Which implies 8 bits per channel (24 bit for RGB) will be plenty - and 8 bit TIFF sufficient?

Or is there some Dynamic Range stretching occurring during the scanning process (that I've never considered before)?

I can see that during/after processing, you might need more bit depth, due to multiplication of values to avoid posterisation, but for capture alone, I'm not sure I see the benefit here either.

e.g. if the scan produces a 'pint', it will fit a 'pint' jug, a 'quart' pot, or a 'gallon' drum, so why use the drum? Surely it just wastes storage space.

Thanks, Dave

Hi Dave,

I know this is an old thread but I just read it yesterday and I feel that I must address the misconceptions that
a)scanning a low dynamic range image at 8 bit is sufficient and
b)converting an 8 bit image to 16 bit for processing will prevent posterization from happening.

When you scan at 8 bit, you effectively throw away most of the usefull data in the image. Let me take an hypotetical old faded B & W photo as an example. The dynamic range will be low. The darkest point might be 100 and the brightest, let say, 200. If scanned at 8 bit, this photo would have a total of 101 levels that we could then stretch over the full range of 0 to 255. Further processing it at 16 bit will not increase the number of levels at our disposal and therefore will not decrease the amount of posterization in the image. On the other hand, if scanned at 16 bit, you end up with 25856 (101 x 256) levels to start with which provide considerably smoother results.

Hope this clarification is helpful.

André

[Dave Humphries]

~ Let me take an hypotetical old faded B & W photo as an example. The dynamic range will be low. The darkest point might be 100 and the brightest, let say, 200. If scanned at 8 bit, this photo would have a total of 101 levels that we could then stretch over the full range of 0 to 255. Further processing it at 16 bit will not increase the number of levels at our disposal and therefore will not decrease the amount of posterization in the image. On the other hand, if scanned at 16 bit, you end up with 25856 (101 x 256) levels to start with which provide considerably smoother results.

Hope this clarification is helpful.

It is André,

I guess an analogy might be that you can have an 8 bit staircase between 1st and second floors, or a 16 bit one. The distance between floors doesn't change, but the step height decreases, by the time you get to 24 or 32 bit, it would be so smooth you'd be able to comfortably come down on wheels

Thanks Dave

PS Thanks also to Stig for the updates this year.

PPS Please - no-one should try to put gamma in to my staircase analogy