Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/jni/affinder

Quickly find the affine matrix mapping one image to another using manual correspondence points annotation
https://github.com/jni/affinder

Last synced: 19 days ago
JSON representation

Quickly find the affine matrix mapping one image to another using manual correspondence points annotation

Awesome Lists containing this project

README 

        
# Description



This GUI plugin allows you to quickly find the affine matrix mapping

one image to another using manual correspondence points annotation.



More simply, this plugin allows you to select corresponding points

on an image, and a second image you wish to transform. It computes 

the requisite transformation matrix using Affine Transform, Euclidean Transform, 

or Similarity Transform, and performs this transformation on the

moving image, aligning it to the reference image.



https://user-images.githubusercontent.com/17995243/120086403-f1d0b300-c121-11eb-8000-a44a2ac54339.mp4



# Who is This For?



This is a simple plugin which can be used on any 2D images, provided

they can be loaded as layers into napari. The images need not be the same

file format and this plugin also works with labels layers.



No prior understanding of the transformation methods is required, as

they perform in the background based on the reference points selected.



# How to Guide



You will need a combination of two or more 2D image and/or labels layers 

loaded into napari. Once you have installed affinder, you can find it in

the dock widgets menu.



![Affinder widget in the Plugins->Add Dock Widget menu](https://i.imgur.com/w7MCXQy.png)



The first two dropdown boxes will be populated with the layers currently

loaded into napari. Select a layer to use as reference, and another to

transform.



![Dropdowns allow you to select the reference and moving layers](https://i.imgur.com/Tdbm1sX.png)



Next, you can select the transformation model to use (affine is selected by default

and is the least rigid transformation of those available). See [below](#transformation-models) for a

description of the different models.



Finally, you can optionally select a path to a text file for saving out the

resulting transformation matrix.



When you click Start, affinder will add two points layers to napari. 

The plugin will also bring your reference image in focus, and its associated points

layer. You can then start adding reference points by clicking on your image.



![Adding reference points to layer](https://i.imgur.com/WPzNtyy.png)



Once three points are added, affinder will switch focus to the moving image,

and you should then proceed to select the corresponding three points.



![Adding corresponding points to newly focused layer](https://i.imgur.com/JVZCvmp.png)



affinder will immediately transform the moving image to align the points you've

selected when you add your third corresponding point to your moving image.



![The moving image is transformed once three points are added](https://i.imgur.com/NTne9fj.png)



From there, you can continue iteratively adding points until you 

are happy with the alignment. Affinder will switch focus between

reference and moving image with each point.



Click Finish to exit affinder.



## Transformation Models



There are three transformation models available for use with affinder.

They are listed here in order of increasing rigidity in the types of

transforms they will allow. The eponymous Affine Transform is the 

least rigid and is the default choice.



- [**Affine Transform**](https://en.wikipedia.org/wiki/Affine_transformation): 

the least rigid transformation, it preserves

lines and parallelism, but not necessarily distance and angles. Translation,

scaling, similarity, reflection, rotation and shearing are all valid

affine transformations.



- [**Similarity Transform**](https://en.wikipedia.org/wiki/Similarity_(geometry)): 

this is a "shape preserving" transformation, producing objects which are 

geometrically similar. Translation, rotation, reflection and uniform scaling are 

valid similarity transforms. Shearing is not.



- [**Euclidean Transform**](https://en.wikipedia.org/wiki/Rigid_transformation):

Also known as a rigid transformation, this transform preserves the Euclidean

distance between each pair of points on the image. This includes rotation,

translation and reflection but not scaling or shearing.



# Getting Help



If you find a bug with affinder, or would like support with using it, please raise an

issue on the [GitHub repository](https://github.com/jni/affinder).



# How to Cite



Many plugins may be used in the course of published (or publishable) research, as well as

during conference talks and other public facing events. If you'd like to be cited in

a particular format, or have a DOI you'd like used, you should provide that information here.