Old maps and great size posters

Discussion forum for techniques and issues relating to the creation of panoramic and/or "mosaic" images

Moderator: spamkiller

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Old maps and great size posters

Post by microfilme » Thu Dec 15, 2011 5:13 am

I work in the Archive of the City of Zaragoza (Spain).
We have done some testing with Autopano.
36 photographs CANON EOS 5D MarkII 200mm lens. The illumination was performed with a Bowens Gemini 400 flash as moving pictures and were made with a photometer measuring the exposure was equal to the above. The images at 16 bits.
Them together very well, correcting distortions, had some failures in joints that we are correcting some hand ...
The software seems good but very expensive to buy. We hope PTAssembler try. Any tutorial or thread for this type panoramic images?
Attached a photo of the old plane on which we are working. 3,5mx1,98m (11,48ftx6,49ft).
We had to "invent" a system to place these documents so great.
We can not build a system to move the document to keep the camera steady. For the tests performed so far we think that the rotation of the camera gives good results ...

Sorry to use an automatic translator.
You do not have the required permissions to view the files attached to this post.

terje.mathisen
Posts: 370
Joined: Thu Nov 27, 2003 6:34 pm
Location: Oslo, Norway
Contact:

Re: Old maps and great size posters

Post by terje.mathisen » Fri Dec 16, 2011 7:17 am

PTAsm should do very well on such a setup, particularly if you can keep the camera still, just rotating it around the proper pivot point for each exposure.

Set the target projection to Rectilinear and watch a perfect scan coming out the other end. :-)

Terje

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Re: Old maps and great size posters

Post by microfilme » Fri Dec 16, 2011 5:52 pm

Thanks Terje.

We have had to make catches in the following way.
9 photos per row. In all four rows.
36 photographs have been the target of 200mm.
I'll try PTasambler did any tutorial that can help?
Luis

terje.mathisen
Posts: 370
Joined: Thu Nov 27, 2003 6:34 pm
Location: Oslo, Norway
Contact:

Re: Old maps and great size posters

Post by terje.mathisen » Sat Dec 17, 2011 3:31 pm

I'd be happy to do a trial stitch for you, just send me copies of the images and I'll return both the PTasm project file and post the image I get out from it.

Terje

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Re: Old maps and great size posters

Post by microfilme » Wed Dec 21, 2011 7:28 am

I send images by email.
The result we will post to the internet for everyone to see what was achieved ...
Thanks.

terje.mathisen
Posts: 370
Joined: Thu Nov 27, 2003 6:34 pm
Location: Oslo, Norway
Contact:

Re: Old maps and great size posters

Post by terje.mathisen » Wed Dec 21, 2011 11:56 am

I have received the first 650 MB of tiff files (jpg would have been better!), it is quite obvious that automatic control point picking does not work.

This means that I have to spend some time manually locating points before I can seriously optimize the stitch.

BTW, I get 200 mm as the lens size. What is the multipier? I.e. is this full-frame, 1.5/1.6 x or from a medium format camera?

Terje

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Re: Old maps and great size posters

Post by microfilme » Thu Dec 22, 2011 4:03 am

It is full frame.
Jpg ¿is better? Let me surprised ...
Were you able to download megaupload files? The second group of photos in TIFF format you send them this afternoon from my house. The internet connection in the office is not very good ...

Have you received email jpg files? You told me.

terje.mathisen
Posts: 370
Joined: Thu Nov 27, 2003 6:34 pm
Location: Oslo, Norway
Contact:

Re: Old maps and great size posters

Post by terje.mathisen » Thu Dec 22, 2011 11:21 am

I am still working on the first 18 images, after defining a bunch of manual control points I do get the optimizer to finish, but without the very good match I was expecting:

A 200 mm lens on a 5D should give about 10 degrees FOV, with very little distortion, but I end up with a (b) value of less than 1%, and control point errors in the 3-6 range, which is about 3X more than I'd like.

Automatic blending with enblend crashes (out of memory; I have 8GB RAM on a 64-bit Win7 installation),so I've fallen back on multilayer PSB.

I'll post some full-size snippets either today or tomorrow.

Re jpeg and the second set of images: Megaupload fails 5 out of 6 attempts on letting me download that 650 MB file of the first images, with JPEG I could have generated a project file which you could have used, after manually replacing the .jpeg file extensions with .tiff.

So far, I have not received any jpegs...

Terje

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Re: Old maps and great size posters

Post by microfilme » Thu Dec 22, 2011 12:13 pm

Yesterday I sent by email (four items) jpg files. I see you have not been reached. Now I can not but then again sending the jpg and now I leave up to megaupload the other group of 18 photos.
A greeting.

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Re: Old maps and great size posters

Post by microfilme » Thu Dec 22, 2011 2:38 pm

I just send your email below to download the 36 files.
Tomorrow I will try to send the work Autopano ... but I try to download the file size obtained = 5 Gb!

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Re: Old maps and great size posters

Post by microfilme » Fri Dec 23, 2011 7:27 pm

In this link
http://www.megaupload.com/?d=TMS3DBHS
you can download the map created with autopano. I had to manually correct 20 errors.
It is a map of 1861 that was created originally by gluing the pieces on a cloth map.
Attention because the areas are "misaligned" are wrong in the original map. Not a software bug.
Sorry to use an automatic translator.

terje.mathisen
Posts: 370
Joined: Thu Nov 27, 2003 6:34 pm
Location: Oslo, Norway
Contact:

First PTAsm results

Post by terje.mathisen » Sat Dec 24, 2011 6:42 am

The SIFT finders did very badly on this setup, so I've had to define more or less all of the control points manually, which means that there are still far less than I would have liked.

I did lose quite a bit of time before I realized that the original map has been patched and stitched together in many places, leading to overlap glitches that look exactly like badly aligned images!

Anyway, here's the first complete stitch, with no manual fixups:

Image

In order to avoid "out of memory" errors from Enblend, I've been working with an output size of just 8881 x 5415, while the default size is nearly 9 times larger: 25990 x 15832 or about 400 Mpix.

My output image is here (11 MB):

http://norloff.org/pano/p.jpg

The image capture has some severe problems though:

a) The camera has NOT been mounted on a properly adjusted pano head, i.e. there are parallax errors. (Possibly also some variation in distance?)

b) Lighting is NOT constant, it seems like the lamp(s) have been moved between shots. One image in particular is a lot brighter then all the surrounding images, but Enblend handled this very nicely.

c) Since the map isn't flat, it has probably been hard to avoid some glare, but several of the images could have given more detail with better lighting.

I'll make one more round of control point insertions and re-optimization and see if I can get a (near) fullsize image out of Enblend.

A question for Max: Is Smartblend better for these large stitches?

Terje

maxlyons
Posts: 3649
Joined: Fri Jun 20, 2003 8:55 pm
Location: USA
Contact:

Re: First PTAsm results

Post by maxlyons » Sat Dec 24, 2011 9:35 am

Terje,

Nice work! The result looks very good despite the difficult input images.

I have been able to use Smartblend to create images up to about 1 gigapixel in size with no problem (on a Windows 7/64 bit system with 8GB of RAM). I haven't used Enblend in quite a while, so I'm not sure what limitations it imposes, but I was under the impression that it ought to be able to easily handle images as large as 400 megapixels. I wonder if the error messages you have seen are misleading, and the real problem is something else (like inadequate disk space in the temporary directory, or some other reason)?

One other technique that can be very helpful for stitching images like this (where the subject is flat, and the camera wasn't rotated around the nodal point) is to use the camera position parameters (CPP) available in the latest beta version of PTAssembler. In theory, you should be able to get a perfect (or very close) stitch using these parameters, but getting the optimizer to converge on a good solution can be tricky. One approach that works quite well is to run the optimizer without using CPP first, and then run the optimizer a second time after adding the CPP for every other image in the project. This usually helps the optimizer to converge on a good solution, although it can sometimes be unstable (converging to a "bad" solution).

Max

microfilme
Posts: 9
Joined: Thu Dec 15, 2011 3:46 am

Re: Old maps and great size posters

Post by microfilme » Sat Dec 24, 2011 11:31 am

Terje great job. Unfortunately that means we are limited. Now we have the head Panosaurus 2.0 and want to have shortly a few diffusers for the flash. With all this I hope to improve photos taken.
On the other hand I read that to photograph a flat object (as the map) the nodal point of aim was not important because there are no problems of perspectives. Is that correct?
Now I would like to learn to manage PTAssambler to avoid having to "bother" people who are so kind ...
Any advice how to start?
THANKS!

terje.mathisen
Posts: 370
Joined: Thu Nov 27, 2003 6:34 pm
Location: Oslo, Norway
Contact:

Re: Old maps and great size posters

Post by terje.mathisen » Sun Dec 25, 2011 7:49 am

If you have a NN or similar properpano head, you should do very well in the future!

The problem with the images I got was that both the distance to the map and the camera position changes a little bit between each image.

As it is I have just finished my final full stitch, using SmartBlend to give me a final image of 24005x15682 pixels.

Having entered 211 manual control points (and fixed one mis-placed point), I'm getting reasonable but not perfect results.

There is just one obvious error, on the left edge, probably caused by moving the camera between the first and the second row, but this should be easy to fix.

I'll send the PTAsm project file so you can redo this yourself!

Terje

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest