How to register the edges of an image mosaic?

[Nasgul]

What would be the best VI to use to register two image edges left/right? I have IMAQ and am willing to overlook parallax errors, just want to get it close and crop out the duplicate image region.

Message Edited by Nasgul on 06-25-2008 01:20 PM

[muks]

What do u mean by register?You can use a clamp get its output cluster and use it to crop up.I u first need to identify the object and then exract it use a pattern matching or co-ordinate system.

Even a line profile across the edges will give u left and right edges reference points.Is this what what u want?

[Nasgul]

I want the viewing pane to display the two images as a single panoramic image, with as little overlap as possible. I'll give the clamp VI a try.

Message Edited by Nasgul on 06-26-2008 06:50 AM

[muks]

IF image stiching is what you want then follow these steps.

Convert first image to array ----> use a reshape array function to convert the array to 1d-----> connect this output to build array

do the same to the second image.In the end DONT FORGET TO RIGHT CLICK THE BUILD ARRAY AND CHECK CONCATENATE ARRAYS.


Now convert this back to images.



Hope i have helped u a lil bit.Let me know If this is not what you want

[Nasgul]

Currently have the edges butted together, but I'm looking to find the matching "seams" in the image and merge the frames so they form one seamless image (the scene overlaps from frame to frame and is somewhat skewed). There's a program called PTGui that does this for mosaic images, but not in real time.

[muks]

Can you post a few images?Have you tried my suggestion?Is the output of trying my suggestion skewed?

---

Dont Forget to rate a good answer

---

[BruceAmmons]

If you don't need realtime, there are much better ways to do it than LabVIEW.  There are probably dozens of photo stitch programs that will do it seamlessly.

However, you say you need to do it on a live image.  A few questions:

Will the images always be at the same distance from the cameras?  Are the cameras fixed in position relative to each other and the object of interest?

If so, you could determine the adjustment of the images beforehand using some sort of template.  This could be manually or you could use corrolation to do it automatically.  During live acquisition, you would just need to do a quick rotate (if necessary), trim, and append.

Bruce