Microsoft Research Image Composite Editor
This free utility that runs on 32 and 64 bit machines has no size limits for image compositing. Although the site says that multiple point perspectives are not currently supported, a geographer from the French equivalent of the USGS says that he has used it to do a multiple view mosaic. The code is NOT opensource, but this could save time IF a community decides that they need the data quickly.
The tool is here: http://research.microsoft.com/en-us/um/redmond/groups/ivm/ice/
If anyone has experience with this, please post as there is nothing on the googlegroup with Microsoft's ICE. I will try this and post results as soon as I get back to the US to a computer with some data sets on it.
PDF version
Comments
Adam,
Microsoft ICE seems to work okay for stitching some sets of vertical mapping photos. It has four stitching modes: one for a camera rotating in the same place, and three it calls "planar." I assume these are for images of something flat, and not necessarily taken from the same place. The planar modes definitely work on map images sometimes, but some sets seem not to stitch at all. I have wondered whether the failures are because some of the images are rotated 180 degrees, and ICE doesn't do that much work trying to find the matches.
Usually the first thing I do with a new set of mapping images is try to stitch some of them in ICE. There is a comparison here http://fastie.net/?p=1241 of a map made in MapKnitter with a version stitched in ICE (with extra images added in Photoshop). It would be really nice to submit stitched images like the ICE version to the PLOTS archive so they end up in Google Earth instead of the aligned but not blended versions from MapKnitter.
Chris
Thanks, Chris. This is great info. Have you ever tried making a composite in ICE and then importing that into MapKnitter to generate a GeoTIFF?
I recall that MapKnitter limits uploaded images to 10 MB, so that would be just a few images stitched together. I have noticed that some MapKnitter maps seem to have the images blended together, so there is more functionality there than I know about. I hope to learn more.
So I just tried a batch of images here from our campus where I work and it solved the composite with only one obvious error in the middle of the road. It looks great and the resulting jpeg is 6.5 MB, but when imported into MapKnitter, the resulting resolution is only 8 cm per pixel. That might be correct, based on the downsampling that appears to have occurred, but it should be higher given that the jpeg has all the data and the balloon was only 500 feet off the ground, tops.
http://mapknitter.org/map/view/wcu-fountain
I'll ask Jeff his thought about this MapKnitter 10 MB question.
And of course there is the obvious issue of the black edges in the map.
MapKnitter apparently accepts png files. I don't know if there is a size limit on png images, but they can have transparent backgrounds. There is one at your wcu-fountain map. It should be straightforward to remove the black background in photoshop and save a jpg as a png, but I'm not sure this is really a solution.
the transparent PNG background feature was broken when we added edge-blending. I hope we can figure out an easy way to resuscitate that feature though, it was nice.
I just noticed that MS ICE will save files as PNG and also as Photoshop. When saved as Photoshop, the background is transparent. So it is easy to get rid of the black background.
The stitched image can also be saved as a Deep Zoom tileset.
Also the three planar modes of stitching are 1) Rigid Scale, 2) Affine, and 3) Homography.
The most serious problem with making maps with ICE is that the stitched result is not georectified and may require extensive rubbersheeting to align with a basemap. Maybe one of the planar modes will be better than others for reducing this problem.
Post new comment