I've been archiving some old photographs from the 1960s with a decrepit Hewlett - Packard Scanjet 8300.
Supposedly, it is capable of 4800 x 4800 DPI resolution, but I haven't seen any reviews on it (like those on Filmscanner dot info) which utilize USAF resolution charts.
Results have been less than stellar (that being said, I don't even have SilverFast software installed ).
I am cognizant of hardware's "effective resolution" - Epson's V850 (6400 DPI) and their latest model, the 13000XL (2400x4800 DPI but with an autofocusing sensor) - have similar effective resolutions.
What I'm curious about is:
1) Why the trend seems to be toward sensors of less resolution, instead of pairing autofocus with the higher - resolution sensor.
2) I am under the impression that "overscanning" at ridiculously high DPIs would provide more "meat" for image reconstruction from learned patterns (meaning, A.I.) in the future.
P.S. Let's ignore the possibility of Epson discontinuing large - format archival flatbed scanners for now.
I've been archiving some old photographs from the 1960s with a decrepit Hewlett - Packard Scanjet 8300.
Supposedly, it is capable of 4800 x 4800 DPI resolution, but I haven't seen any reviews on it (like those on Filmscanner dot info) which utilize USAF resolution charts.
Results have been less than stellar (that being said, I don't even have SilverFast software installed ).
I am cognizant of hardware's "effective resolution" - Epson's V850 (6400 DPI) and their latest model, the 13000XL (2400x4800 DPI but with an autofocusing sensor) - have similar effective resolutions.
What I'm curious about is:
1) Why the trend seems to be toward sensors of less resolution, instead of pairing autofocus with the higher - resolution sensor.
2) I am under the impression that "overscanning" at ridiculously high DPIs would provide more "meat" for image reconstruction from learned patterns (meaning, A.I.) in the future.
P.S. Let's ignore the possibility of Epson discontinuing large-format archival flatbed scanners for now....Show more →
Here are my thoughts
On question #2, overscanning as you call it can be helpful in some situations. The media your are scanning needs to contain that level of detail and also fine tone level variations for the higher resolution to be of value.
One question #1. Scanning companies are in the business to make money. They need to make a profit after recovering all R&D, Manufacturing, and Marketing/Sales costs. These days, demand for high-end scanners is quite low, making such models a poor business proposition. The sales volume is already low, and pricing the scanner higher to make it more profitable would likely reduce sales volume and, in turn, revenue for that product.
So the scanning companies are trying to hit the sweet spot of features/specs and costs ot sell in a reasonable volume.
Technically, it is quite possible to create the more advanced scanner you refer to, but it apparently is not a good business proposition.
You might benefit from a colour profile for your scanner, but that depends a lot on what you are scanning.
Resolution is important but a good scanner needs to have a good tonal response too, for all colour channels. Poorer scanners don't scan as much of the colour and tonal range as the better scanners. Colour profiling generally helps but only if the colours are actually seen and distinguished from each other by the scanner.
Scanning slides/film is quite different from scanning prints. The former is more likely to be adversely affected by curvature of the item being scanned, and that will undermine the benefits of resolution. The latter would not need very high scanner resolution but the colour range may be more restricted. Autofocus is likely to be done once per item rather than for every part of the item. Pro outfits would use drum scanners instead of flatbed scanners.
Scanning modern digital prints at very high dpi is even more troublesome because you'll end up with lots of almost-floral patterns that each replaced a single original "pixel" when the photo was printed. These were ok when printed at 300-600dpi but look silly when scaled up to bigger sizes after scanning.
So, in general, scanning media at very high dpi tends to create excessive amounts of artificial / incorrect / misleading data rather than useful data, because there just wasn't enough data in the source material. Sure you can fix it up to some degree, but that is generally achieved by effectively reducing scanned data rather than revealing more of what the original picture might have represented.
Separate from that issue is that scanning well takes time. Lots of time. The higher the resolution the slower the scan and the more time it takes to scan. At this point you haven't even seen the scan file let alone processed it. And, very importantly, it takes a long time to ensure that you don't lose or shuffle your original media so that you'll be able to find it again when working on a scan reveals that it justifies a rescan and more effort. Are all your boxes of originals identified? uniquely? Can you you actually find what you want again later on? quickly enough? If you shuffle/re-sort them then you may lose useful info about the date of capture or brand of film, etc.that might only be on the packet; it won't be in your scan files.
My somewhat limited experience suggested I would average about 10 minutes per picture to get the scanning "right" without losing track of the originals. That's an awful lot of time to process thousands of items. And "special" pictures will still want further editing at some point.
I would give silverfast software a look, it might provide better control of the the scanner. (Scanner manufacturers typically don't invest in updating scanner software) And if your scanning film slides you can pay a small fee to silverfast and unlock 16 bit color.
there are ways to convert a digital camera to scan slides as well as photos. The digital camera really put a dent in scanner sales. The peak of scanner sales was during the transition from film to digital. Now it seems there are very limited choices for high end scanners
Regarding " over scanning" a quick google came up with the following:
Scanning at a higher-than-optical resolution does not provide additional useful information for AI photo reconstruction. AI reconstruction models are trained to infer and add plausible high-frequency details from the data that is already present. If the original scan lacks detail, the AI will create, or "hallucinate," the missing information rather than retrieve it from the image.
once your scanning past optical your scanner is generating information the AI programs can do better.
It also takes much longer during the scanning process.
I haven't seen it discussed much around here, but Vuescan may be worthy of your consideration. For years, Ed Hamrick has gone above an beyond in reverse engineering scanner function and supplying an almost constantly updated scan program capable of driving almost any known scanner, flatbed or film. Vuescan has the capability of working with known targets, reflective and transmissive, for color and density calibration.
You can download it for an unlimited test drive, with the understanding that all scans will have an overlaid watermark.
FWIW, I have absolutely zero connection other than being a long-time, satisfied user.
I've been archiving some old photographs from the 1960s with a decrepit Hewlett - Packard Scanjet 8300.
Supposedly, it is capable of 4800 x 4800 DPI resolution, but I haven't seen any reviews on it (like those on Filmscanner dot info) which utilize USAF resolution charts.
Results have been less than stellar (that being said, I don't even have SilverFast software installed ).
I am cognizant of hardware's "effective resolution" - Epson's V850 (6400 DPI) and their latest model, the 13000XL (2400x4800 DPI but with an autofocusing sensor) - have similar effective resolutions.
What I'm curious about is:
1) Why the trend seems to be toward sensors of less resolution, instead of pairing autofocus with the higher - resolution sensor.
2) I am under the impression that "overscanning" at ridiculously high DPIs would provide more "meat" for image reconstruction from learned patterns (meaning, A.I.) in the future.
P.S. Let's ignore the possibility of Epson discontinuing large - format archival flatbed scanners for now....Show more →
What are you scanning?
My response is based upon your interest in producing higher quality scans. I’ve ignored your technical questions, because I don’t think that it gets you where you want to go. Today, your least expensive approach to high quality scanning is using a digital camera. Without spending a small fortune, this won’t quite get you to top end film scanner levels, but close. The scanner you are using will not give you great scans and moving to an Epson won’t greatly improve the results. Many B&W films and some colors, like the old Kodachrome slide film will give you results approaching 40-50MB in 35mm, 70-80MB in 645 and 90-100MB in 6x7. Theoretical maximum resolutions far exceed these numbers with Kodachrome 25.
I’ve scanned with multiple Nikon Coolscan scanners including the 9000ED. Also used the Epson 700 with aftermarket film holders. Only an expensive flatbed would exceed the Nikons. The flatbeds are faster and can provide a contact sheet. I have the same project underway where I’m scanning thousands of images from 35 mm to 6 x 7. I think the idea of producing contact sheets with the flatbed and then specifically archiving the better images is a good approach and it allows you to use slower, but higher resolution scanners for the images that you really want.
I think that you can pretty much ignore the resolution claims made by HP, or Epson.
I never purchased a DSLR, much less a newer mirrorless camera and lens, after my ancient Memory Stick Sony Cybershot.
I once did some picture - to - film salvaging of old photos with an SLR, it was a tedious experience for a few pictures (mainly because I only had an inappropriate lens). I imagine it would be easier with a digital mirrorless + proper macro lens and some additional equipment - but is this a painful task for say, twenty rolls of 36 - shot film and approximately 800 family album - sized pictures?
With my current line of work, having a flatbed scanner at hand seems more beneficial - and adding a contemporary $500 Plustek + $500 software seems the least painful financial insult.
Alan4color wrote:
I would give silverfast software a look, it might provide better control of the the scanner. (Scanner manufacturers typically don't invest in updating scanner software) And if your scanning film slides you can pay a small fee to silverfast and unlock 16 bit color.
Presently thinking about it.
Would you think my present scanner’s just too old, though?
Lousy older - tech CCD, something like that?
Alan4color wrote:
Scanning at a higher-than-optical resolution does not provide additional useful information for AI photo reconstruction. AI reconstruction models are trained to infer and add plausible high-frequency details from the data that is already present. If the original scan lacks detail, the AI will create, or "hallucinate," the missing information rather than retrieve it from the image.
I’m presently scanning at 600 DPI, which produces 8MB PNG photos on the average (not even sure that’s the best format - just chose it because it was lossless and these images are to be viewed on digital devices anyway). The scanner is quoted to do 4800x4800 DPI optical, so in reality that would be 2000x2000 DPI at the most (just guessing, based on what I read at Filmscanner dot info, seems to be reasonable to reduce stated optical values in half).
What I’m curious about is going at a higher optical resolution, not a higher softeware interpolation (which is in itself primitive AI) - would this give future AI more meat to work on?
Alan4color wrote:
I would give silverfast software a look, it might provide better control of the the scanner.
---------------------------------------------
Taperwing wrote:
I haven't seen it discussed much around here, but Vuescan may be worthy of your consideration.
---------------------------------------------
If I were to "upgrade" in an attempt to get better results, I should go with the software first, then the flatbed scanner (if need be), then the Plustek.
Does that sound like a reasonable stepwise incremental approach?
EDIT: Just to put things into context, my scanner is twenty years old. Cost $500 then, was supposed to be an upper - end model.
).