I am curious - is there a lightweight solution that detects edges (of a document) in a video frame / image that doesn’t depend on OpenCV (which is way too large)? Something like CIDetector (with CIFeatureTypeRectangle) on iOS?

OpenCV is a huge library but I think you can compile it with only the needed functions. What do you think about building it as a static lib? I am working on a similar problem with OpenCV (but not for edge detection) and I'm curious about your approach.

Unfortunately I don’t have any approach and haven’t worked with the NDK yet. I’m currently trying to evaluate if this (using Renderscript) can be extended: https://www.zacsweers.dev/detecting-blurriness-with-renderscript/ But thanks for the hint. It’s really a shame that Apple offers so much more regarding image processing.