The Gui is very helpful and works really well as a tool for evaluating intensity based image registration. However the core function is a convergence algorithm and possibly because of this, never seems to produce pixel perfect or near perfect registration. It's main advantage is obviously to automate the process without requiring control point user intervention.
Conversely from my experience, imtransform seems to produce better registration results and I can get pixel perfect registration in most cases for my application, but this requires user intervention for control point generation.
What would be really good, from my viewpoint is an algorithm which takes the initial control points from the base image and then automatically searches for equivalent matches in the unregistered image. Any help you can provide in this direction would be a definite five star!
I am not sure what you mean by "manual...GUI." But it probably wouldn't make any difference if I did. While IMREGCONFIG, IMREGISTER, and IMREF3D all facilitate 3D image registration, I haven't incorporated that capacity into this application, nor am I aware of any GUI that does. You might find the demo on "Registering Multimodal 3-D Medical Images" useful, though.
Eric, "it doesn't work properly" doesn't help me fix it. I use this frequently in both R2013a and R2013b, and without problems. If you can steer me to the issues you're finding, perhaps I can get to the bottom of them.
And the inability to generate the tform reflects the current status of IMREGISTER, and isn't necessarily a limitation of this tool. That said, function IMREGTFORM was added to the Image Processing Toolbox in R2013a, and my updated version reflects that change. (And provides the tform as an output.)
Thanks for the rating,
It doesn't work properly in 2013a and 2013b, in that there doesn't seem to be a way to have it generate the t-form matrix.... Otherwise, it works great in older versions! I was bit harsh in the last rating...