Debate: camera positioning -- flexible vs. fixed

[jerrytsai]

The positioning of the cameras may be 1) fixed; 2) capable of constrained movement (somewhat flexible); or 3) capable of movement in all directions (completely flexible).

Examples:
Fixed (mostly)
> Ultem's hovering platen build: mounted to platen
> Rob's inverted platen build: mounted, can be linearly moved, with hassle, along the side of the platen at intervals

Somewhat flexible
> daniel_reetz's The New "Standard Scanner" : mostly fixed, but capable of rotation around a bolt
> gsloop's New build...: adjustable mount with linear movement

Completely flexible
> clemd973 pioneered the completely flexible approach in Clemd's DIY build in photos by using gooseneck microphone extensions.

The question is: how much of a problem is inflexible camera positioning?. I would imagine that an inability to, parallel to the page, move the camera toward the spine of a book would result in keystoning. One might also be less able to avoid unwanted reflections, although that might be solvable with different light positioning.

On the other hand, the prevalence of fixed or somewhat fixed camera positioning in many builds implies that lack of flexibility in camera position is not a great drawback.

Some thoughts:
Might platen angle matter (e.g., 90 degrees vs. more obtuse angles)?

Complete flexibility in positioning allows one to position a camera closer or farther from the page. This could be worthwhile for those interested in maximizing the resolution of the image.

NB: clemd973's solution does comes at a cost. I estimate it costs about $36 for a pair of extensions, adapters, and straps versus about $6 (assuming the 2x4 wood pieces are "free" as sunk costs) for the New "Standard Scanner".

Please weigh in. In your build, has flexibility in camera positioning mattered? Is flexibility worth paying for?

[Drake Ravensmith]

I haven't done much serious scanning yet but I have the same setup clemd has and I can tell you that it's a little frustrating. Getting the image just right is proving to be difficult because of the micro adjustments involved. I've also wondered if a fixed mount might not be easier but I don't have the interest in setting one up at the moment.

[daniel_reetz]

Great question. The basic bottom line is this: if your camera has enough pixels to capture the entire platen area at your target resolution, then adjustability is not important. Just nail the placement once and be done with it. This will work fine for smaller platen sizes.

However, if you want to move the camera to get more resolution on smaller format books, it might pay to design a mount that can accommodate the camera in two fixed positions, one up close, and one further away.

[jerrytsai]

I would imagine, for the most rectangular (i.e., least keystoned) image, the lens of a camera should be positioned over the center of a page. So being able to move in a plane parallel to a page (or, analogously, parallel to the platen pressing down on that page) would be helpful. But if keystoning is not a detriment, however, then having this ability is not all that important. So I suppose I am asking the forum this question: just how bad is keystoning? I would imagine the less severe the keystoning the better. The closer an image is to flat, the less software would have to compensate.

Moving nearer or farther from a page to maximize the resolution of the images is probably NOT that important a function in that cameras can zoom. Whether to create two mounts, one near and one far, to ensure maximization of resolution depends on how important resolution is to the one who is scanning. I have read a post in which someone was trying to ensure a minimum of 300 dpi (dots per inch). Others have mentioned trying to achieve 600 dpi. I would imagine this problem to disappear as technology improves with time and even inexpensive point-and-shoot cameras will pack more megapixels than anybody needs.

[the.traveller]

My suggestion is to have several points for the camera to mount.

To keep the cameralens middle (imaginary crosshair) pointed towards the book diagonal center you need a flexible mounting position.
The bookpage has an 45 degree angle, so the camera needs to be fixed in an 45 degree angle to ensure that there is no (skewing/keystoning?).

So I suggest measuring the different book sizes, then measure the diagonal center of the books. (Just cut a paper the same size as the book, take a ruler draw lines diagonally) Then you will know how high the center is from the bottom of the horizontal plane of the base. Imagine a straight line (or take a long stick which I do) from that center upwards. Then you will see how high the camera mount should be.

The reason for taking the diagonal center of the books is that lenses tend to get more (blurred/out of focus, whatever)
towards the outside. You will have the best quality in the middle. That is the reason why you don't get raw pictures from compact camera's. By slizing the edges of the picture, you won't notice it isn't that sharp on the edges as in the middle.

So for different sizes of books you need different heights where the camera can be mounted to insure you take pictures centered on the diagonal middle of the book.

Now since the "New standerd build" has 2 columns which are fixed on a corner, there is still the problem that your cameralens is not pointed to the book center, so you will get a border somewhere around the edges of your book.
So to keep the cameralens centered towards the diagonal center of the book one should not only have the possibility to adjust the camera in height (z-axes) but also vertically (y-axes).
Because the camera has the ability to zoom, this will make sure the picture will be filled as much as possible with a page from the book. (x-axes)

So the angle your camera in which your camera needs to be mounted is always the same as the angle of the cradle
(total cradle angle degree -(1/2 * total cradle angle degree) )
(i.e. 90-(1/2 * 90 = 45))
(120-(1/2*120=60))

So when you are building a cradle which is a 120 degree slope the camera needs to be higher on the z-axes to have the same angle. Some houses (cellars/atticks) don't have that high ceiling.

Then there is the problem of how large your cradle is. If you want to make the cradle as large as your largest book to be scanned but also your paperbacks then you will have the problem that you can not lower the cameramount below the height of the cradle.

Only solution for that is to have different sizes cradles and platens.

-------
Addition 11-7-11
This morning I have found an article explaining much better, with pictures, what I am trying to say.
The only thing which is not discussed is the vertical (y-axes).
Fix Keystoning via hardware adjustments

[snaguy]

Fixed but flexible to the platen. My Book Scanner has the platen attached to 4 vertical drawer sliders. The camera also attached to the drawer sliders with an adjustable CCTV mount. This eliminates the need for micro adjustments throughout scanning a book. Once the setup has been initially calibrated/adjusted there should be no problems completing a book. I think a lot of people are wanting external screens because of movement throughout the scanning process. The book liberator style is fixed but with the below mount adjustable. The parts I use in my scanner would also suit a book liberator style which I will make later once I have tweaked my current setup. My scanner does have some issues with the weight of the platen and lifting. I am sure I will come up with a nice solution. Also some reflection issues which also I am not to worried about solving.

My current scanner is a mid range. A Book Liberator style would be a budget. If I ever get the time and funds I will make a high end with linear slides and Arduino controlled automation.

What type of books are you looking at scanning?

CCTV Mount available on eBay AU $6.58

ux_a10110300ux0155_ux_c.jpg (17.44 KiB) Viewed 2008 times

My Book Scanner
http://diybookscanner.org/forum/viewtopic.php?f=1&t=1069

Book Liberator Link
http://bkrpr.org/doku.php

[daniel_reetz]

Here are the two problems with keystoning.

1. If the page image is keystoned a bit, the resolution across the page is not uniform. Meaning that the smaller end of the trapezoid has less pixels to represent each bit of the page.

2. Keystoning also means that the page image is not normal to the camera. That means that one part of the keystoned page will be focused slightly better than another part. For the most part, this is not an issue with compact cameras, because they have deep DoF (depth of area that is properly focused), but it can be with cameras with larger sensors, which have shallow DoF.
[sometimes this is multiplicative because the smaller end of the trapezoid is also out of focus]


For both cases, the issue is dependent on your intent with scanning. Many people want to end up with a Scan Tailor binarized/black and white output, and they want to read that directly. In those cases, both are not really a problem. However, if you want to see the page as it was, with the image of the paper, or you are running OCR on your images, it is important to have as little keystoning as you can. That doesn't mean that you need to eliminate it entirely, but it does mean that you should take steps to reduce it as much as you can afford in your time, skills, and intent budget.