I don't know a lot about Phonon, but I think if you create a MediaObject type to serve your frames off the disk, then you will be able to use the framework. It looks like the Effect type is what you would use to implement frame-by-frame processing. My understanding from reading the documentation is that Phonon uses what they call a "graph" (actually a processing pipeline) to move frames from a media source (like a file) to a media sink (like a video widget). You can insert nodes into this pipeline, branch it, connect multiple inputs together, etc. Looks pretty powerful.
If you don't have audio associated with your images, you might be able to get by using QMovie with a custom QImageReader class. If you designed the QImageReader class correctly, you might be able to chain them together to implement your on-the-fly processing.
Marking things on screen means that you will want to save that information somewhere, right? That means that you might have time-stamped metadata associated with your images, so that when you play it back the second time, the metadata marks appear along with the video. I think in that case you would have to use something like Phonon, and use an Effect to merge the metadata with the image frames.
Sounds like a hard project, but interesting. The book "Advanced Qt Programming" by Mark Summerfield has a chapter on QMovie and Phonon with example code.
Bookmarks