Czur, project review for 2016

I don't work for Czurtek, but I contributed to their Indiegogo project and received an ET16 document or book scanner early in 2016. Here is a few thoughts on how my perception of this project has changed.

First, I don't want to make this a long article, but wanted to pull together my shifting thoughts on this project here on January 15, 2017.

When I started with the scanner I had thoughts it would be similar to a Aztec Booksnap or Fujitsu 4220c or somewhere inbetween. I was wrong.

It does scanning well and produces a collection of scan images [per session] and then provides a [work table] on which you can perform clean-up of the images individually or in bulk, then choose to create OCR - Optical Character Recognition, and bind the touched up images and the OCR information into a Searchable PDF file for archiving or reading later.

That was the dream, but it fell short in a few areas.

The Desktop PC for Windows software had some challenges in communicating what it could do, how you should use it, and in how to update it.

The Desktop PC for MacOSX software was not released and its been quite a while, but a new pre-release demo now exists which appears very similar and very good.. but so far also lacks good documentation.

The Desktop PC for Linux (if it is planned) has also not been sighted.

Over the last year I have had good luck at contacting support and they have responded with a private Twain32 driver for Windows, and an updated Twain32 driver for Windows (I have yet to review).

They have also shown me an SDK for developing programs on Windows that by-pass the Desktop software and scan images direct to the file system.

On my own I've come to understand that they are not very precise on releasing details about the chipsets or optics used in the hardware, and my guess is that this is due to Non-Disclosure Agreements with iCatchtek, or SunPlus which seem to do a good job of keeping their datasheets and programming guides off the Internet.

But on the good side, it seems the resolution and optics are so good as to put precisely considering and managing those details are not as important as they once were. From an engineering or obsessive detail persepective this can be frustrating.. but in the end really not that important. Any optical barrel distortion effects are confined to such a small field of view its just not that important.

This is also not that much different from the level of detail available for the Apple iPhone.

Its important to understand too that USB 2.0 or USB 3.0 has limits in the available bandwidth for capturing an image and moving it over to the PC across USB. Too long a delay is simply not acceptable, so often MJPEG or Motion JPEG of a high resolution is used to moved scan images across the USB cable. YUY2 (uncompressed raw) is available, but only at lower resolutions.

The controls on the desktop software are more or less like the [Auto] or coarse level controls available on 'Point and Shoot' pocket Cameras, demanding few specifics. These limited Profiles set the scanner camera, lights and image capture chip up based on the general question [what are we scanning today?] by way of -- 1. Is it a Full Color Photo? 2. Is it Color Line Art? 3. Is it a B&W Photo? 4. Is it Simple B&W Text?

Considerations like Dots Per Inch, Filters for Moire, and other things are left to "post processing" after the image is acquired.

On the PC the post processing is carried out using the OpenCV libraries.

As best I can tell the OpenCV libraries for the PC running Windows, use the DirectShow interface to create a FilterGraph, which is an object that continuously grabs frames from the scanner and passes them into a Null render object. On its was to Null however is a general purpose [grabber] object which can copy a single frame and place it in a memory buffer setup by OpenCV to receive the image, then any of the functions in OpenCV can be used to "Post process" the image in preview to optimize what will be scanned and captured then post processed into an image file on the PC.

Since the original scanned image is also available, before processing it can be saved too. This is as close to raw as we can get for now.. but its really not that important.

The major features touted for "this" scanner are the ability to "automatically" straighten or flatten the image, based on the additional ability to take a lower resolution scan with Laser beam guides. The high-res image and the low-res image are then "stretched" in OpenCV to mask and correct the 3D distortion from bending or folding until the Laser bean guidelines are straight. Quite a marvelous item in one combined product if all goes well. Some people call this "De-warping" the image.

 I am not sure how, but original images can also seemingly be re-examined and re-post processed, which suggest the lo-res image may be embedded into the original hi-res image file for possible later use.. but I could be wrong and just not understand the system.

The long term goal however is to make scanning and binding into an electronic document or eBook (in the form of a PDF) easy and without much thought. I think this would have worked.. if details about how it was handling the images were provided, and demos included clear instructions from A to Z on how to do this.

While certainly possible on Windows with the Desktop PC software, until recently it hasn't even been possible on the MacOS with the Desktop PC software.

The cloud solution seems to have trouble that users outside China experience as long unacceptable delays requiring an always connect state with the Internet.

The scanner is partially UVC compatible, meaning default programs on Windows, MacOS and Linux identify it, load their drivers and will use it for taking pictures, but generally low resolution pictures, and the default programs do not 1. provide fine control over the image contrast and brightness or colors 2. provide 2D straightening or 3D flattening 3. provide automatic OCR 4. provide a work table for touch up or binding into a PDF document.

In summary my perspective and expectations have changed over the year.

This is an Amazing product, still in development, and will be a great value when it is complete. And the renewed release of new firmware and Desktop software, the release of Desktop software for the Mac, release of Twain32 and release of a second generation of Twain32 driver plus an SDK that includes the ability to 2D straighten and 3D flatten.. all point to a very bright future for this project and product -- but -- it is not finished!

The biggest challenges are the "language" barrier and lack of sufficient documentation to not only "achieve" document scanning and binding.. but that guides and teaches.. and makes you comfortable with it.. are its biggest challenges today.

I regularly receive a lot of offline commentary and inquiry about this scanner and requests for my opinions.. but I always preface those conversations with 'I do not work for Czurtek..' so I cannot and do not speak for them.

I can say I wouldn't mind serving in some way to help with the documentation, but I think the language barrier flows two-ways.. I do not think they always understand what I am saying when suggesting small improvements that would make a 'big' difference in how people perceive their product.

Luckily I have a full time job, so the only benefit I get from making videos or answering questions about the Czur ET16 is when I learn something for myself or someone politely says a video or article helped them. And for that I thank anyone who happens to read these articles..it motivates me to keep going and not give up on the ET16..and sometimes I do feel like giving up.

But with patience and time, I can see its potential.. and that really there is nothing currently on the market like it. Its super simple, its complete, the results are really good.. and its still being improved.

Usually when something comes to market, all development stops and its frozen in feature time.

This scanner is not.. its improving and its getting better.