And I am well aware of how the compression works by having i-frames that hold all information of the one image, while P and B frames save file space by only holding reference information that is based on the i-frame. Your knowledge is correct with regards to how mp4 and mpeg2 work. If you remove the I frame there is nothing to reference subsequent frames to. An I frame is sent that is the complete frame, subsequent intermediate frames are simply a code saying, the same as the last frame except for this bit, this bit, and this bit, then when the next I frame comes along, it can compare the guesses made on previous frames and confirm they are correct. I may be completely wrong as I never work with mp4 files, only DV-AVI and DVD compliant mpeg2, but will you even have anything at all if you remove the I frames? Assuming mp4 is a better compression than mpeg2 but works in the same way, the I frames are the only ones that contain the full frame.
