Align multidimensional images in 2D or 3D, rigidly or with warping, whatever the difference in scale. Provide also a set of tools for Correlative Light Electron Microscopy, but can be used for registration/alignement of any correlative imaging.
Contact us for any questions, or post in the image.sc forum.
This plugin allows to compute a similarity (translation/rotation/scaling and flipping) transform from pair of points. It is updating the transformed image interactively such that the user get immediate feedback. The transformation is saved and can be applied to any other stack/image. Non rigid deformation can also be applied in 2D or 3D.
Please make sure to use the last version of the plugin (update plugins in Icy). Source code available on GitHub.
3D/3D,2D/3D or 3D /2D can be handled . 3D ROI are enabled, and can be checked with the 3D vtk view (size of ROI can be changed using the ROI stroke width). Some prelalignment by rotating in 3D the volume is possible.
Transformations can be applied directly or combined through Block Protocols (search for apply transformation).
It’s also provide information about the predicted Error (based on statistical prediction as described by Fitzpatrick et al), either as a full color mapping, either on each points used as landmarks, and error on the discrepancy in position between points.
Authored by Perrine Paul-Gilloteaux and Xavier Heiligenstein.
List of video tutorials:
Registration of a slice to a volume (2D to 3D) (same principle for 3D to 2D, just inverse source and target)
Pre registration by 3D volume rotation from 3D viewer.
Reapply a previously computed transform (tutorial in 2D, but exact same procedure for 3D)
Integrated preprocessing (such as flatening options or denoising)
List of pdf tutorials:
First step for manual 2D-2D rigid registration:
Press Play to initialize placement of points.
For 2D: place points on target image, on image source drag it on the correct position. After 3 points, images will be automatically updated when added a new point. Transfirmations are saved in a file named after the source file suffixed transfo.xml
For 3D: Same principle: it is advised to use the duplicate view capacity of ICY to follow the pacement of points in 3D. For now, make sure to go to the slice of source image containing the landmarks you will add in target image. After 4 points, stacks will be automatically updated.
For both 2 D and 3D, if you select the let me update myself option: the transformation will be computed only once you pressed Update Transformation. You can add more points and click again “Update the transformation” to update the transformation.
The non rigid option work also based on landmarks, in 2D or in 3D directlly. Note that the image will be warped such that the landmark are now in the same position. Press
“In want to prealign (rotate in 3D)” will allow that you turn the 3D view (VTK Canvas) and then generate a transformation saved as xml as well and a new stack in the interactively indiacted view (useful to flip a 3D volume for example).
Apply transformation can also be apply in a protocol (in you combined a 3D prealignement (_MANUAL TRANSFO.xml), with a 3D RIGID transformation, followed by a NON RIGID deformation,) , you can create an Icy protocol to cascade these transformations with the Apply Transformation block.
- 2D option: will compute a transform ignoring the Z position of the landmarkes points, but it can be applied to multidimensional dataset.
- 3D option: can be used to compute 3D to 3D, 2D to 3D (placing a reoriented slice in a 3D volume), 3D to 2D (extracting a reoriented 2D slice from a volume)
- Non rigid: use with caution at it will warp your image (an infinity of solutions are possible here…) It is advsed that you use a rigid transform first: EC CLEM will advise you to go or not to a non rigid deformation, based upon the difference between an expected error after rigid registration, and the actual error observed on used landmarks.