Scaling scans of analog photo negatives

Questions and postings pertaining to the usage of ImageMagick regardless of the interface. This includes the command-line utilities, as well as the C and C++ APIs. Usage questions are like "How do I use ImageMagick to create drop shadows?".
Post Reply
JK1974
Posts: 5
Joined: 2013-03-22T08:07:21-07:00
Authentication code: 6789

Scaling scans of analog photo negatives

Post by JK1974 » 2013-04-03T08:52:11-07:00

Hi all,

I am about to scan photo negatives with VueScan from Hamrick Software. But I don´t want to simply convert the negatives to positives and store them as downscaled and heavily compressed JPEGs, I first want to save high-quality digital equivalents to the analog negatives in the way described at http://www.colorneg.com/scanning_slides ... re/VueScan, building a kind of "digital archive", mainly because digital algorithms e.g. for dust removal and (automatic) color correction become better from year to year while the quality of the original negatives degrade at the same time. I don´t want to do any (heavy) post-processing on this archive. The images are saved as "RAW TIFFs" (RGB and a fourth alpha channel that contains dust and scratches information taken from an infrared scan; 64 bit = 16 bit for each channel).
Inverting the colors, removing dust and scratches, maybe even further scaling should be done in a second step - and could be repeated later easily using batch processing if any of the algorithm gets better over the years.

Even though I still have some problems opening the generated RAW TIFF images with ImageMagick (see viewtopic.php?f=3&t=22037), I am facing the problem that the images are too big - without any benefit. Every image currently has a file size of ~125 MByte with an image width of around 5000 pixels. The reason why I use this high scanning resolution can be found e.g. here: http://www.filmscanner.info/en/PlustekO ... 7600i.html:
If a scanner with a nominal resolution of 7200 dpi provides in practice a resolution of only 3250 dpi, then it is obvious to scan with half of the resolution of 3600 dpi. But a resolution test with 3600 dpi resolution resulted in only 2600 dpi of effective resolution provided by the scanner, thus approximately 20% less than in the case of a scan of 7200 dpi...
I compared a scan with 7200dpi against a scan with 3600dpi and did nearly see no improvements in details (at least on my photos), so I stayed with 3600dpi in the scan settings (I don´t believe in the mentioned 2600dpi - but this is a different story). Using lower scan settings clearly showed a decrease in details.
But if I want to get the "real" resolution and compare the resulting 3600dpi scan against scales (done in Photoshop) with the resolutions of 8 and 10 Mpix digital cameras (3456x2304px/3872x2592px), I believe to see a loss in details in 8 Mpix resolution, while 10 Mpix seems to be ok. If I scale to this size of 3872x2592 pixels with Photoshop, I get a file size of around 75 MByte/image - which is far better and IMHO no more waste of space.

But then I face the next problem: Which is the best downscaling algorithm for this task? Photoshop´s bilinear and bicubic are known to be outdated, that´s one of the reasons why I want to use ImageMagick.
I have read the posts in this forum about downscaling and the impressive summary at http://www.imagemagick.org/Usage/filter/nicolas, but in my case, there is a different situation: I don´t have the colorspace of positive images and not the sharpness of digital images, but of course, I also don´t want Moiré (I don´t expect this at this level of details but you never know) and want to avoid any sharpening/ringing artifacts at this "archive" stage. Especially the alpha channel that carries the dust and scratches information (infrared channel) should not have any ringing artifacts because it is the base for future dust and scratch removal algorithms.

Do you have any recommendations, ideas or experiences on this special case?

Thanks a lot in advance!

henrywho
Posts: 188
Joined: 2011-08-17T06:46:40-07:00
Authentication code: 8675308

Re: Scaling scans of analog photo negatives

Post by henrywho » 2013-04-03T09:11:09-07:00

any sample image? :?

snibgo
Posts: 11992
Joined: 2010-01-23T23:01:33-07:00
Authentication code: 1151
Location: England, UK

Re: Scaling scans of analog photo negatives

Post by snibgo » 2013-04-03T09:13:09-07:00

Regarding file sizes, are you using a "-compress" option? You might find that this significantly reduces file size without losing any data (at the expense of processing time). Eg:

Code: Select all

convert in.tiff -compress Zip out.tiff
snibgo's IM pages: im.snibgo.com

JK1974
Posts: 5
Joined: 2013-03-22T08:07:21-07:00
Authentication code: 6789

Re: Scaling scans of analog photo negatives

Post by JK1974 » 2013-04-03T16:21:28-07:00

I currently can´t test the -compress option because the alpha channel is being changed to a transparency channel in Photoshop after having opened the file with ImageMagick (see the link in my first post about the problem/bug opening the RAW TIFFs). However, if this bug is not being fixed in one of the next versions, I want to try to find a different solution, e.g. splitting the alpha channel from the RGB channels and store it as a separte tiff etc.
Nevertheless, I still think that storing the data in this resolution is a huge waste of space, even if I use the -compress parameter - unless your recommendation is that scaling might be a no-go for my "archiving purposes". I want to digitize the content in the best (affordable) quality for private purposes before problems like color degradation occur. That´s why rendering times also are no problem in my case (as long as scaling does not take 10 minutes per image... :D).

This is a (temporary) download link for one of my scans (original unaltered VueScan RAW TIFF output file, still as negative): https://dl.dropbox.com/u/17946310/Minol ... FC400-8.7z
I made this photo not long ago (in 2005!) with a Minolta 7000 AF (USA: Minolta Maxxum 7000 AF, Japan: Minolta α-7000) camera, using a Kodak FC400-8 negativ film. The image shows some thin lines on boats in the foreground that might help judging the level of sharpness (however, you will also note a lot of image noise), while the roofs of the houses in the background might help to determine any Moiré effects when scaling. Unfortunately, I currently can´t remember a photo with brick buildings or similar structures that might be suited better, but maybe you get at least a first impression.

snibgo
Posts: 11992
Joined: 2010-01-23T23:01:33-07:00
Authentication code: 1151
Location: England, UK

Re: Scaling scans of analog photo negatives

Post by snibgo » 2013-04-03T17:32:38-07:00

In the UK, a 4TB external disk costs around £130. If an image takes 0.2 GB, the disk would hold 20,000 images. So less than a penny per image, and the cost keeps dropping. Double or treble that for backups, of course. But still less than the cost of negative sleeves and binders.

Scaling algorithms keep improving, like everything else. See various discussions at viewforum.php?f=22 . Personally, I don't reduce the resolution of archive images. I don't know how I might want to use the image (or a pertion of it) in the future.
snibgo's IM pages: im.snibgo.com

JK1974
Posts: 5
Joined: 2013-03-22T08:07:21-07:00
Authentication code: 6789

Re: Scaling scans of analog photo negatives

Post by JK1974 » 2013-04-03T23:43:37-07:00

Using an external disk as backup seems to be common sense, but I have already lost 2-3 external hdds in the last years (luckily without important data). One of them was an internal Samsung disk that got too hot in an external case (I have another two Samsung disks from the same series at that time in different PCs that also started making noise when it had not been cooled with an extra fan), and then I had problems with disks that did not spin up again, and then there could also be the classic: the headcrash. The first cases don´t make that recovery impossible, but expensive.
Because of these experiences, I still intend to archive the data on DVD. (Nearly) all of my CD-Rs and DVD+/-Rs of the last nearly 20 years are still readable (apart from one charge of low-price media), furthermore, there are tools like dvdisaster that I could use to create recovery data that might help in case a disc would partly become read errors. However, with a size of 125 MByte each, I in fact would have to use an additional compression to make 36/37 images fit on one DVD-R media. And I don´t think that I have more than 50 negative films from 2000 to 2006 that I want to digitize.

However, I have not thought about using a RAID system. This might then also store precious video data.

Post Reply