Topic: Computed Tomography resolution increase along Z axis from FIB-SEM

Hi folks,

This is a little off topic but I hope someone can help.  I am a student studying a PhD in Nuclear Materials Characterisation.  One of the avenues of research is using a focused ion beam to cut samples in slices, collect an image of each slice with a scanning electron microscope and reconstruct the tomographic stack into a 3D model which we can conduct experiments on using a physics simulation package.

There is one problem with this.  I am currently cutting slices 50 nm apart, my resolution per pixel on the X and Y axis is roughly 3 nm.  Therefore I have X and Y axis pixels of a high resolution and Z axis pixels of a low resolution.  This makes my voxels elongated and cuboidal.  In order to even up this situation I need to add extra images to fill in that 50 nm gap between slices.  I have used a blend tool but this fades between images ghosting features between slices.  This isn't representative of my samples.  I need something that will predict the change between images more intelligently.

This is where frame interpolation comes in.  If I am able to use frame interpolation to predict the change between my slices then I would produce a much more representative set of filler images to generate a 3D model.  Is it possible to use a stack of .tiff file images with this software?  Could I create an animation of the images I have, increase the frame rate using this software and then extract the frames as image files?  Does anyone know a good tool to interpolate more than two images together (I have stacks of roughly 200 to 300 images)?

I know this isn't the usual use for this sort of software but I do feel like it could be extremely useful to the scientific community if it could be applied.

Many thanks for your consideration.

2 Reply by dlr5668 (edited by dlr5668 19-06-2019 11:23:30)

Re: Computed Tomography resolution increase along Z axis from FIB-SEM

ffmpeg https://stackoverflow.com/questions/249 … ith-ffmpeg to loseless video -> process with svp -> split back into images

you can also buy 2060 and try https://developer.nvidia.com/opticalflow-sdk

Re: Computed Tomography resolution increase along Z axis from FIB-SEM

I already knew how to convert my images to a video format, I was just wondering if this software had an output including the interpolated frames or if it was done on the fly.

The Nvidia optical flow is extremely interesting though.  We currently have an older generation Quadro in our modelling computer but I am looking to upgrade that and I have had my eye on the Quadro RTX 8000.  This would be another reason to go down that route.

Thank you very much for your reply!

4 Reply by dlr5668 (edited by dlr5668 19-06-2019 15:14:10)

Re: Computed Tomography resolution increase along Z axis from FIB-SEM

ben.veater wrote:

I already knew how to convert my images to a video format, I was just wondering if this software had an output including the interpolated frames or if it was done on the fly.

The Nvidia optical flow is extremely interesting though.  We currently have an older generation Quadro in our modelling computer but I am looking to upgrade that and I have had my eye on the Quadro RTX 8000.  This would be another reason to go down that route.

Thank you very much for your reply!

https://www.svp-team.com/wiki/Manual:SVPcode
Example u have 1 fps video. Use 2x rate and it will provide 2 fps video. SVP can use nvidia optical flow as well btw

Re: Computed Tomography resolution increase along Z axis from FIB-SEM

dlr5668.  Thank you so much for your help and input with this.  I will go away and have a good think about my options.  You've been extremely helpful.