Posted by
Kenneth Sloan-2 on
May 23, 2018; 8:33pm
URL: http://imagej.273.s1.nabble.com/How-to-align-images-manually-and-merge-them-into-a-multi-channel-composite-image-tp5020699p5020720.html
With which tool?
The one I recommended (align with line ROI) does translation, uniform scale, and rotation.
Most of the ones I saw suggested will offer a choice of possible transformations, usually including:
*translation only
*translation and rotation
*translation, rotation, and uniform scaling
*and more...up to arbitrary rubber-sheet warping
I had a rather nice (but crude) version that allows you to manually specify 1, 2, or 3 landmark pairs, and
then selects the type of transformation based on your choice. I was about to make it bullet-proof, when
I discovered "align with line ROI", which does everything I need (short of automatic registration) at the moment.
The major problem I have in automatic registration is that my images tend to be multi-modal, meaning that the shapes and the
intensity values are radically different from image to image. Things like SIFT work great...until they don't.
Since I'm lazy, and I only write the code, and not use it - I'm not unhappy pushing the task off on the poor image reader.
My current guideline is that I will use auto methods if:
a) the imaging is fairly consistent, with the major difference being the position of the camera wrt the scene, or
b) there are more than 20 images in a given stack to be aligned (in these cases, I try to process away the
imaging differences so that the registration plugin can handle it.
Otherwise: my current standard is "align with line ROI". Most of my "customers" spend a long time examining
the images, so the extra time to locate 2 landmarks (in each image) is not onerous. Your milage may vary.
--
Kenneth Sloan
[hidden email]
Vision is the art of seeing what is invisible to others.
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html