Author Topic: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI  (Read 1887 times)

0 Members and 1 Guest are viewing this topic.

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #25 on: 2017 Jan 07, 07:12:09 PM »
Super-resolution will work with scanner images also, but the key to making it work is moving the sensor or subject by at least half a pixel's width for each image. Basically, super-resolution digs out all those half-pixels to turn 1 pixel into 4 pixels.  So, you can clearly see that 400% resolution gain is not hard to achieve. Amazing.

So, i should run a 2400dpi and 4800dpi image of the same coin through this..
  • never argue with idiots
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #26 on: 2017 Jan 07, 07:44:01 PM »
I tried running PhotoAcute on 2400dpi and 4800dpi images, but the result had colors all messed up, what settings did you use for the image above? Also choosing the camera profile is just a guess.. I'll try the other apps you listed..
  • never argue with idiots
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #27 on: 2017 Jan 07, 08:11:04 PM »
You probably tried to correct chromatic aberration. Uncheck that box.
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #28 on: 2017 Jan 07, 08:15:00 PM »
It says Scanning Resolution: 2400 dpi (main scan), 4800 dpi with Micro Step (sub scan)

My 4800 dpi images look much better than the 2400 dpi scans [...]

The hardware might be doing super-resolution. There are some cameras that do it too, but they don't call it superresolution, they call it their own proprietary names. If it is doing super-resolution, then that's honest image data at 4800 DPI. Normally they would call it "interpolated" if it were not honest data. Or, another possiblity is that it's just an advanced upsizing method that doesn't actually add any new image data. That's what ReShade does - just advanced upsizing.
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #29 on: 2017 Jan 07, 08:16:10 PM »
I split off a post that broke the topic display again:

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI

You can try uploading the problem files here:

https://vgy.me/

Then maybe I can submit the files for examination by the bug team for our forum software.
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #30 on: 2017 Jan 07, 08:19:57 PM »
I split off a post that broke the topic display again:

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI

You can try uploading the problem files here:

https://vgy.me/

Then maybe I can submit the files for examination by the bug team for our forum software.

Yep, i can't upload anything anymore, not even 2400dpi jpeg's, which worked yesterday.

I tried uploading a scan of the classical garden antique. Playing with reshade atm, converting a 4800 dpi BMP to a 9600 dpi jpeg.

judging by the 'micro-step' language in the scanner specs, i'm assuming it can do optical 4800 dpi by halving the servo steps
  • never argue with idiots
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #31 on: 2017 Jan 07, 09:03:03 PM »
Well, my initial findings playing with Photoacute and Reshade:

I can't get Photoacute to work at all, the colors are all messed up, and the picture gets cropped etc. Photoacute is designed to combine 2 slightly different images, so what i should do is take a scan, then microscopically nudge the coin a bit and take another scan. I've tried the different presets and removing chromatic etc.. no luck so far, but again, the software isn't designed for this.

With Reshade i tried supersizing a 4800 dpi image to 9600 dpi, and zooming into the result vs the original showed almost no difference at all.

Based on my experience so far i get the best images with 4800 DPI period. The 'Unsharpen Mask' filer in the scanning software works pretty well, and is needed, because the coin is behind plastic etc, and so doesn't lie directly against the glass bed, where the focus is.

Anyways, this method is better than taking pictures with a camera for documentation purposes at least, It's much easier to scan a large number of coins, the results are very comparable between each other because of the controlled, similar lighting conditions for each scan. Also i don't have to wait for the right kind of daylight etc.

I'll take 4800 DPI scans as BMP of all my coins now, and hopefully later we can get the upload issues fixed, so i can put them all on display, and then proceed with scanning my non-mcc collection as well.

I'll upload some scans of the classical garden antique and later silver to the address badon gave, i'll post a 2400 DPI scan, a 4800 DPI scan with and without sharpening and backlight correction, so that you can maybe compare them side by side and give feedback on the best method to settle on. I'd like to find good settings before scanning all the coins in one go later.
  • never argue with idiots
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #32 on: 2017 Jan 07, 09:56:43 PM »
i'm uploading scans of the Antique Administrator's Garden. In 2400 dpi and 4800 dpi, with and without Unsharpen Mask and Backlight correction filter. Also 9600 dpi scans as reference. The images above 4800 dpi are made by scanner software, because FastStone can't convert them from BMP. The rest are converted from BMP with FastStone.

Badon maybe you could compare them a bit based on description in filenames and give your input. Personally, i'm settling on 4800 dpi with medium Sharpening filter to offset the coin being slightly out of focus. No backlight correction or other filter. 4800 dpi and sharpening only.

Still holding back on the most exciting scan, which will be the 2016 Nanjing copper PF70 of course :)

https://mega.nz/#!jAISSC4Z!kKU_Owb2btWrqejb9CboOlc9Z7rjd0-n0J2TfHFJ8-Q
« Last Edit: 2017 Jan 07, 10:11:46 PM by Y7ASyxC »
  • never argue with idiots
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #33 on: 2017 Jan 07, 10:30:26 PM »
Keep the original, unaltered BMP. Don't do any sharpening or other manipulations to it. Preserve the original, and verify data integrity with Corz Checksum. Then use MultiPar to generate recovery data to correct errors in the future. Use Corz Checksum for quick verification, so you never need the MultiPar data unless an error is detected.
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #34 on: 2017 Jan 07, 10:36:15 PM »
Keep the original, unaltered BMP. Don't do any sharpening or other manipulations to it. Preserve the original, and verify data integrity with Corz Checksum. Then use MultiPar to generate recovery data to correct errors in the future. Use Corz Checksum for quick verification, so you never need the MultiPar data unless an error is detected.

Yes i will always scan an original without any modifications. Badon could you still compare the 2400 and 4800 DPI images and give your opinion about if it's 'true' 4800 dpi or not?

I use the Zettabyte filesystem extensively, which mitigates any need for 3rd party crc checking really, but then again, it's still needed when using the files on other systems. Thx.
  • never argue with idiots
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #35 on: 2017 Jan 07, 10:43:45 PM »
Actually, i now see there is no difference on the images regarding sharpness, seems it only concerns scanner software image preview. Plain 4800 dpi it is then..
  • never argue with idiots
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #36 on: 2017 Jan 07, 10:46:06 PM »
Keep the original, unaltered BMP. Don't do any sharpening or other manipulations to it. Preserve the original, and verify data integrity with Corz Checksum. Then use MultiPar to generate recovery data to correct errors in the future. Use Corz Checksum for quick verification, so you never need the MultiPar data unless an error is detected.

[...]

I use the Zettabyte filesystem extensively, which mitigates any need for 3rd party crc checking really, but then again, it's still needed when using the files on other systems. Thx.

This forum and the CC runs on ZFS too. It's still possible for errors to occur. I always use at least 2 different ways to detect and correct errors. ZFS can detect errors, but it's not always checking for errors, and it has limited data recovery ability. We have to use a 3-way ZFS mirror to ensure that there is almost always going to be at least 1 mirror with accurate data. If you're not using a 3-way mirror, you're highly vulnerable to data loss. For static data, there is no good reason to not use MultiPar. It can correct errors even if they go undetected for many years, including user errors, which ZFS has no ability to stop you from doing.

I still use multiple methods to verify data copies and transfers, and correct errors when they occur. MultiPar is not CRC. Corz Checksum is not CRC.
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #37 on: 2017 Jan 07, 10:47:35 PM »
Can you give me a link to the images you want me to compare?
 

Offline Y7ASyxC

  • never argue with idiots
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #39 on: 2017 Jan 07, 11:36:00 PM »
I'm having issues with a silver coin, the reflection is too bright. I'm also starting to wonder if that strong light on the scanner can 'darken' the surface of silver coins?
  • never argue with idiots
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #40 on: 2017 Jan 07, 11:52:18 PM »
If you have a polarized filter, it's possible to filter out some of the reflections, but I'm not sure how well that will on a scanner bed. By the way, the ultimate high resolution photography is done with a scanning back to medium format camera. Basically, it's a flatbed scanner with a lens in front of it. Maybe if you're curious, you can make some kind of abomination out of your scanner. Maybe try it with a pinhole lens and see what happens :)

Don't worry, light alone normally has no effect on silver. If there's heat, that might cause toning to happen faster, though. Heat can also activate sweat glands in your hands that can cause problems with silver.
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #41 on: 2017 Jan 08, 12:04:06 AM »
Well it is what it is, i'm now scanning clean 4800dpi images of all coins, the rest can be done with image manipulation software if needed.

Here is the problem coin, original, and darkened.

https://mega.nz/#!aEpAnJTa!E8qNHKdhjTDxZmUt2iq6o6hvUm0kArh_Wd6-xOqCktM
  • never argue with idiots
 

Offline Y7ASyxC

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #42 on: 2017 Jan 08, 12:40:29 AM »
Badon, all the jpg images in the Antique and silver Garden .zip file are converted from BMP using FastStone with default jpeg settings. So you can test uploading those images if you are looking for the cause

The copper coins are looking pretty nice :)
  • never argue with idiots
 

Offline badon

Re: Flatbed Scanner test - 2017 G500Y from 1200 DPI to 9600 DPI
« Reply #43 on: 2017 Jan 08, 03:48:37 AM »
I just examined your 2400 and 4800 DPI images, and the 4800 DPI images do have a little more detail, but not enough to justify a 350% to 400% increase in file size. I'm not 100% sure the scanner isn't cheating.

For example, camera aficionados rave about Fujifilm's cameras, but in my tests, I discovered the camera was cheating to give the impression of higher quality when in reality, there was no extra detail present. They were cheating so bad, I will not buy a Fujifilm camera, even though I like the quality of their lenses. The specific reason I won't buy their cameras is because their cheating actually causes a slight LOSS of legitimate detail.

It's akin to the way ReShade manipulates the data to give the impression of higher resolution, when none actually exists. If you tried to use ReShade upsized images in a super-resolution process, the results would be worse than using the original data. That's why it's so important to preserve the original data as much as possible. Super-resolution weeds out the liars. I had a bunch of articles and demo data prepared to publish. It probably would have made a big splash in the camera world, but I got bored and did something else, haha.