Easy workflow to stack several frame from a steady video with a moving subject

There are plenty of samples of what I want to do: a moving subject in front of a steady camera, and a single final picture showing the “action” with the subject in several positions in front of the fixed backgound.

I want to do this not from a burst of still image, but from selected frames of a video.

The easy way I found (on Windows) is the software Microsoft Image Composite Editor, but it has some drawback:

  1. it is not so fast
  2. it only allow to mask the moving subject (i.e. to select what area of what frame to preserve in the final stitched photo) with a rectangle and we are not able to define that rectangle in an accurate mode;
  3. after viewving the final result, to change the area to keep of a frame or the frame itself we need to rebuild the final stitch.

To solve the (2) we can of course extract the frames from the video and stitch with Photoshop, but the (3) become more difficult.

Any workflow to recommend?