I am working on a blob tracking project and have many high-definition videos that I would like to reduce in size for storage and downstream tracking/shape-analysis. I want to use a lossless method that takes advantage of the black and white nature of the video as well as the fact that not much is moving between individual frames.

The videos are quite sparse, with 5 to 10 b&w blobs per frame occupying <30% of the space in total, with each blob moving <5-10% of the field of view between frames and not changing shape too much between 2-3 frames.

I will work in Python, Matlab, or LabView for this project, and could use a batch utility if available. It may be worthwhile to export the files as compressed image stacks if a proper video format can't be found. What are the pros and cons of this? A video codec uses correlations between neighboring frames, so it should be more efficient, but not if the wrong one is chosen or if it is improperly configured.