From the READMEs I gathered used of xidr and corr2p would give me something I could pass to estpchirp2m's -pin option; However, when I tried this, I still lost. What I'm trying to do is "stitch" pieces of a scanned map back together; sheetfed map scanners cost more than I have, and so basically I have 4 "pieces" of a map with a few hundred pixels of overlap (but not necessarily an even boundary) that I want to stitch.
Not to work against the project which I think is great but what I used to do before the project was started was use GIMP.
Create image large enough for both parts.
Put each part on its own layer.
Use one layer as the background and the other, with a lower opacity, to move as neccesary.
This is how one had to do it before computers.
Again, I do not mean to impede the project but if this is so urgent, here is a mthod.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I'll be working on an improved xidr (probably with james)
this weekend. This should be able to properly obtain
the 8 parameters you need to stitch the images together,
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
You also can seed the estimation with an initial guess.
For instance, lets say the images are about a quarter shifted to the side from each other.
Try "-pin 1.0 0.0 0.0 0.0 1.0 0.25 0.0 0.0"
or for up/down shifts, try
"-pint 1.0 0.0 0.25 0.0 1.0 0.0 0.0 0.0"
Use the -display option to check out whether to use +0.25 or -0.25 for instance (cant remember which way is which offhand). You should see them appear at the right spots when you have the right +/-
Don't forget to use mutliple steps "-steps" option too for best results.
Finally, there's also a program in there called "est_trans_fft" which automatically deals with large image shifts.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
From the READMEs I gathered used of xidr and corr2p would give me something I could pass to estpchirp2m's -pin option; However, when I tried this, I still lost. What I'm trying to do is "stitch" pieces of a scanned map back together; sheetfed map scanners cost more than I have, and so basically I have 4 "pieces" of a map with a few hundred pixels of overlap (but not necessarily an even boundary) that I want to stitch.
I have thus far been unable to hint estpchirp2m in the right direction. Could you provide an example, perhaps, of using it with images with limited overlap? Alternately, the images I've been using for experimentation, if you wish to look, are at
http://prr.dementia.org/temp/cur/alexandria31250/alexandria31250ne.jpg
and
http://prr.dementia.org/temp/cur/alexandria31250/alexandria31250nw.jpg
but just a kick in the right direction with any images would be great, if you have time.
Thanks
Not to work against the project which I think is great but what I used to do before the project was started was use GIMP.
Create image large enough for both parts.
Put each part on its own layer.
Use one layer as the background and the other, with a lower opacity, to move as neccesary.
This is how one had to do it before computers.
Again, I do not mean to impede the project but if this is so urgent, here is a mthod.
I'll be working on an improved xidr (probably with james)
this weekend. This should be able to properly obtain
the 8 parameters you need to stitch the images together,
You also can seed the estimation with an initial guess.
For instance, lets say the images are about a quarter shifted to the side from each other.
Try "-pin 1.0 0.0 0.0 0.0 1.0 0.25 0.0 0.0"
or for up/down shifts, try
"-pint 1.0 0.0 0.25 0.0 1.0 0.0 0.0 0.0"
Use the -display option to check out whether to use +0.25 or -0.25 for instance (cant remember which way is which offhand). You should see them appear at the right spots when you have the right +/-
Don't forget to use mutliple steps "-steps" option too for best results.
Finally, there's also a program in there called "est_trans_fft" which automatically deals with large image shifts.