View on GitHub

Quorten Blog 1

First blog for all Quorten's blog-like writings

So, using a DSLR camera, my Pentax K-1 DSLR body and 1980s 50mm SMC Pentax prime lens to be specific, has been an interesting journey. First of all, I’ve been impressed with the volume of photos I”ve been able to take. I’ve quickly reached over a thousand photos of my documents. Now I want to flip through those pages on a larger computer screen to look at what I have.

Unfortunately, that’s where the first problem comes into play. Although flipping through the 8K Ultra High Definition photos on my DSLR camera works perfectly fine, I’ve noticed problems with my Ubuntu photo viewers responding a bit slowly. Sluggish. So, what is the cause of this problem.

After some more in-depth analysis, I’ve identified and timed these key steps in processing required for photo viewing:

  1. The time duration to load the data from disk.

  2. The time duration to decompress the JPEG image, or to demosaic the RAW image.

  3. The time duration to pan and zoom the loaded image.

I’ve found out that steps #1 and #3 are reasonably performant. Since I already knew in advance that I wanted to page through many photos quickly and didn’t need high color tonality for predominantly black-and-white documents, I’ve only saved directly to JPEG, not using a RAW development workflow. So, #2 was my performance bottleneck: decompressing 8K UHD JPEG images.

There surely are solutions to this by the way of GPU functions: hardware accelerated video decode APIs and GPGPU programming are two such potential solutions. Another possibility is better optimized CPU decoders. I’m going to have to look into this further and provide updates on what I find.

In the meantime, I’ve been looking toward improving other aspects of my document scanning workflow.


I believe this is the first time I wrote this, so I’m not going to claim I’m reiterating something that I didn’t already write before. However, I may have noted this one earlier: having properly sleeved film negatives makes a big difference in how quickly and efficiently you can scan film negatives. With paper documents, this fact is much more extreme. For film negatives, the worst condition they may be in is that they were left in the original developing envelopes that came back from the commercial film developing shop. On the other hand, the worst condition for paper can get far worse: It’s not uncommon for papers to be poorly preserved to the extent that they are bent out of a flap shape, or even creased and crinkled.

So, the same advice to expedited film scanning applies to paper, but in a much more extreme sense: items that are properly packaged for scanning pays off with dramatic time savings and increased efficiency for the person doing the scan operations, ScanOps.

For film negatives, the ideal packaging is as two-fold:

  1. The developed film strips should be placed within high-quality, non-adhesive negative sleeves. Removing a film negative should be as easy as gently slipping it out the side.

  2. The negative sleeves should be inside a box binder. This is a special type of 3-ring (or n-ring) binder that closes such that all the edges are sealed tight from elements of the weather such as dust, rain, and light. Both the “box” and the “binder” elements are important: the box, of course, for weather sealing, and the binder so that the scanner operator can quickly page toward any negative sleeve to select for scanning.

For paper documents, the ideal packaging situation is a little bit more curious to the uninitiated. The actual ideal packaging situation is exactly as you find reams of new printer paper packaged:

  • All papers of the same size are precisely stacked up into a ream of 500 sheets.

  • An easily removable sleeve surrounds the whole stack of papers.

  • The reams are stored horizontally, just as you find multiple reams of new printer paper stored in an office. (Well, maybe only a typical one.) Incidentally, this is also an archival best practice to avoid weird page curl-ups and other physical aberrations that can happen from vertical storage.

So, why are these recommendations in place? First of all, there is the obvious physical reason to prevent the documents from succumbing to the fate of poorly preserved ones: bent, curled, creased, and crinkled sheets of paper that require extensive restoration work to bring back to a reasonable condition. Second, by packaging the papers into the same form factors used by standard office printers when possible, the same paper feed mechanisms that are used in those high-volume office printers can also be used in a high-volume document scanner: an automatic document feed (ADF). Bingo, you’ve just eliminated the toughest part of the scanning job.

However, there are still plenty of other paper-like articles left to be scanned that do not fit into such a nice “digital library” mindest. For bound books, yeah we have automatic page turning robots, but that tight crease in the center still poses a problem for scan quality and speed. A 3D scanning intermediary step can be used to compensate for page curl, but that slows down the overall scanning process. Conversely, the spine can be sheared off a book and the pages run through an automatic document feed, but that effectively destroys the original book. Finally, there are still many more oddball-sized sheets of loose leaf paper left. These generally do not fit through an automatic document feed with any amount of ease.


So, what is your last resort for scanning paper documents that works in all of these situations? Yep, that’s right: turn those pages by hand. After all, that is how you read the documents and books the first time, right? Naturally, all of those paper-based media formats support this mode of reading by design.

The practical limit on turning pages by hand is about one page per second. Faster than this, your hands might reach dangerously fast speeds that can damage the documents you’re trying to preserve. Plus, even at one page per second, the human vision system isn’t fast enough to keep up with what’s going on: you’re hands are often times just a blur to your eyes. In any case, at the end of the day, you realize that scan speeds of 1000 pages per hour or faster make your DSLR camera a real workhorse compared to that old flatbed scanner where you can only scan 6 pages per hour.

The practical limit for scanning film negatives by hand paging is about 100 negatives per hour on average, or about 2 negatives per minute. Again, delicacy is paramount to preservation, so this often times forms your natural speed limit that prevents you from scanning much faster than this.

Finally, remember to factor in the bandwidth limits of your DSLR camera into consideration. Even though you might have the tools and technologies to scan faster, if your camera can’t keep up, you’re going to have to slow the rest of your system down to work with what you have, or else shell out thousands of dollars to get a faster DSLR camera. With today’s (2018) DSLR cameras, the practical speed limit is not the optical systems, but the digital systems: the digital data buses that complete the circuit of writing the data out to storage or an external bus simply can’t write out data any faster. Incidentally, this is also one of the main reasons why DSLR cameras can have nearly identical optical systems, but differ in price by a factor of 10.

That being said, archival photography is generally not a lucrative area, so even if your main interest is in archival photography, you’re better off settling with a DSLR camera where the digital buses are fast enough for typical daily use, even though this may leave room to be desired for archival photography.

For my DSLR camera, the Pentax K-1 camera, the practical limit on SD card write speed is 26 MB/s. Peak write speeds can be up to 32 MB/s under special circumstances. With 20 MB JPEG photos, this means that 1 photo per second is the practical limit. Therefore, there is no speed advantage for me to use an ADF system: hand page turning is just as fast, plus it is more versatile than the ADF that only works under special circumstances.


What does it take to reach a page turning rate of 1 page per second? There are a few practical physical layout requirements to optimize to this speed.

  • Pages should be turned book-style: papers should be stacked up in two piles, and pages should be photographed two at a time.

  • In order to be compatible with loose leaf paper, the stacks must lay horizontal. You can’t place them on an angled or vertical surface, even though this would make setting the camera up more convenient.

    • Consequentially, this means that your camera must either be operated hand-held or supported on a tripod that allows it to look directly down at the floor. Tripod mounting is ideal for speed as it allows your hands and visual concentration to be fully focused on turning the pages quickly.
  • The tops of the stacks should be close to each other in height, so that they are within your lens’ depth of field. For loose leaf papers, this means picking up the papers in smaller packets at a time, and paging through the packet in a book-style manner.

  • Finally, my own self experience. It feels like I’m “in the pit” when I’m working on the scanning stuff: It’s actually physically strenuous to turn pages quickly purely for the sake of photographing them.


So, let’s discuss better tripods. I started out with an ultra-portable lightweight travel tripod that cannot look directly down at the floor. Plus, my DSLR’s camera is just below the weight limit of the tripod. Obviously I need something better for archival photography, and by any means I need the better tripod simply for a higher weight capacity.

So, what’s in stock for this?

  • Professional tripods are purchased separately in two parts: the legs and the ball head.

  • The professional ball heads are serious business: they do not limit your degrees of freedom like the ultra portable cheap ones do and they lock very tightly.

  • The professional tripods have center columns that can swing out so that you can orient your camera directly downward. Also, they have mount points so that you can place counter-weights on the opposite end of the swung-out center column.

  • Expect higher prices for this higher-end equipment. However, if you plan to shoot thousands of photos for many hours in any case, “your back will thank you” for getting the right equipment.

  • An L-plate is great complement to a professional tripod: it allows you to mount your DSLR camera in either portrait or landscape mode. Make sure you get one designed specifically for your camera, not a “universal” one, and do not be fooled by excessively expensive L-plates either, as this is still just a piece of metal.

20180520/DuckDuckGo tripod look directly down
20180520/https://photo.stackexchange.com/questions/13003/how-can-i-take-horizontal-photos-where-the-object-is-lying-flat-on-the-floor
20180520/https://photo.stackexchange.com/questions/7419/what-should-i-look-for-in-a-camera-tripod-for-photographing-microfilm-machines-a
20180520/http://learnmyshot.com/photography-tips-for-shooting-overhead/

This is a great resource.

20180520/https://improvephotography.com/12062/tripods-ball-heads/
20180520/DuckDuckGo tripod gear site:improvephotography.com
20180520/https://improvephotography.com/gear/tripods/
20180520/https://improvephotography.com/42411/l-plate/


I do realize that as the photo subject section in my blog is taking off, I’ve ought to add some of my photos I’ve shot into the posts I write. Alas, I’ve yet to figure out a hosting solution that I’ll like for the photos. Ideally the service would be a “storage-only” one that focuses around making the images accessible through a URL.