I am currently exploring using cell images to make a generative terrain. I had been working on a project last semester which took hyperspectral microscopy data as a source for generating 3d terrain-like meshes. Here are some of the meshes I made:

Here is a more in depth explanation of how I got this data and what I was trying to do with it: http://invizibility.github.io/work/Hyperspectral.html

I want to continue this design study and ideally would like to make a larger scale terrain a user could fly around in. These meshes were created with simple displacement maps, however I want to see how I can extract cells from the images using CV and tile together blobs of cells strategically so as to make an interesting landscape. Here is the tSNE grid organization of the yeast cell data that I have: