Message Boards Message Boards

1
|
1983 Views
|
2 Replies
|
1 Total Likes
View groups...
Share
Share this post:

Unexpected result from ImageAlign on 3D Objects

POSTED BY: J. Leko
2 Replies

Try changing the Method option. The most common method of image registration is to use keypoints, which are sharp locatable features (like 'corners'). The more of these, the better it works. Your start image likely has at least some, but your problem image is very fuzzy. There are other options which might work better.

POSTED BY: Gareth Russell

As suggested, I conducted a (brief) investigation of the other “Method” options (see attached notebook below). The FourierBlurInvariant was a qualified success. While it is able to register the two samples, and keep them within the specified volume, it adjusts the intensity values of the data. The bright yellow spot visible in the day 4 sample is missing from the ImageAligned version.

Note that the KeyPoints method returned errors. I attribute this to the use of 3D volume data rather than 2D images. Implementing these techniques on 3D volumes is also probably raising the complexity level!

Are there any other suggestions to perform image registration on 3D volumes?

Thank you for your consideration.

POSTED BY: J. Leko
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract