User:Jhellingman/Czur Scanner Review
A review of the CZUR Ultra scanner
As a volunteer for Project Gutenberg, I scan a lot of books. Most of the time, I’ve used flatbed scanners, but these are slow and have become increasingly niche—and therefore more expensive—in recent years. (This excludes cheap scanners based on LiDE technology, which are unusable for book scanning due to their total lack of depth of field.)
Another reason to avoid flatbed scanners is that they can be destructive to books. To get a good result, you need to press the book flat, which often damages the spine. Since I have numerous antiquarian books, that is unacceptable. Finally, most common flatbed scanners handle only paper sizes up to the US “Legal” format. This means that many larger books cannot be scanned as a page spread, but instead require scanning each page individually, slowing things down even further.
For this reason, I had been eyeing planetary scanners for quite some time. Unfortunately, these are either very expensive or require a lot of DIY effort. Then the CZUR scanner caught my eye—a relatively inexpensive planetary scanner—and after some experimentation at the dealer, I decided to buy one. Being inexpensive, I was already aware of some of its limitations, but I also realized that the speed gain (as well as preserving the spines of my books) would make it worthwhile.
Since then, I’ve scanned many thousands of pages with it.
The Good
The scanner arrives in a very neat box, which is worth keeping if you want to transport it to books you cannot bring home. All accessories are packed in separate boxes.
- A set of additional lights to improve illumination of scanned books
- A foot-switch, allowing you to hold a book flat with both hands
- An integrated design, so no need to build a DIY solution using a smartphone
- Three red beams projected onto the page, intended for automatic de-curving (more on that later), which also serve as a useful visual indicator that a scan has been taken
- The ability to scan A3-sized books (by scanning left and right pages separately), or A4-sized books two pages at a time
The Bad
- The software that comes with the scanner has a number of usability issues and can hardly keep up with the scanner’s output, even on a 2021 laptop with a Ryzen 5 processor.
- When scanning, I sometimes work too quickly and need to verify that I haven’t skipped a page. To do this, I have to wait for the software to finish processing, then click on a thumbnail to view it at a large enough size for confirmation. It would be much better if the most recently scanned page were displayed more prominently.
- The included USB cable was of such low quality that I had to replace it.
- The finger cots are a nice idea, but not particularly practical.
(I cannot judge the included OCR software, as I have never used it. Instead, I rely on an older (2003) version of ABBYY FineReader 7.0, which still has features essential to my workflow that have mysteriously disappeared from later versions.)
The Ugly
- The quality of scanned illustrations is abysmal. The software applies very strong edge enhancement, which turns illustrations into something resembling comic-book art. While this sharpening may help with text recognition, it makes the scanner unsuitable for artwork. I have also found no way to disable this in the software. To give an impression, I’ve included a cut-out from a scan made by the CZUR scanner (left), and the same cut-out from a scan made on a Canon 9000F flatbed scanner (right).
- The de-curving feature does not work reliably. Although I expected this to be useful, in practice it is simply too problematic to use, and I abandoned it after a few attempts. The software would sometimes rotate pages by 20 or 30 degrees, cutting off text in the process. This often required re-scanning the same page—sometimes with the same flawed result. Worse, it would regularly introduce unexplained breaks in the text, as if there were a small fold in the page, causing narrow columns or diagonal strips of text to disappear. Because this is difficult to detect during scanning, it often only becomes apparent during proofreading after OCR—by which point it is far too late to correct. Again an example of this phenomena, to make it clear:
Since I use the open-source program ScanTailor to post-process my scans—and it handles de-curving better (even without laser guidance)—this limitation is not a major issue for me.
Summary
- For quickly scanning text for OCR: excellent.
- Automatic de-curving is not usable:
- Results are haphazard and often worse than the original scans
- Pages may contain discontinuities where letters are missing, rendering OCR output unusable
- Scanned illustrations are close to useless; I still rely on my flatbed scanner for these
Final Thoughts
For me, the scanning speed alone justified the purchase. However, in many respects, this feels like a product that could have been much better. Given that even inexpensive smartphones now have high-resolution cameras, and that open-source software such as Fairscan can produce very good results, it is clear that this device could achieve significantly more with better software.