pythonopencvimage-processingimagejimage-stabilization

ImageJ: Image Stabilizer


Problem: Stabilize a stack of images

Question: How could I perform image stabilization without key-point detections? Is there any way to do it?

I tried this: http://www.cs.cmu.edu/~kangli/code/Image_Stabilizer.html in ImageJ and it does stabilize. The method used is - Lucas-Kanade. I checked many sources and it all needs key-points (like corners, etc) to be defined. [Source: https://link.springer.com/chapter/10.1007/978-3-642-25975-3_34]

Are there any ideas how I could stabilize the stack of frames without key-points?

Images are microscopic cell images (with cell-division)


Solution

  • If you don't wan't to use keypoints you need to look at all the pixels in the image and use a measure that can compare the similarity of two images, and try to optimize this similarity.

    These measures could be

    etc.

    ElastiX is a tool that supports these two measures and more, and you can allign two images at a time, so you will need to do it sequentially.

    However, you will probably get just as good (and faster) results using keypoints.