4

I've been studying methods in reducing temporal redundancies in video compression and about the main concepts in MPEG video compression. I understand that I-frames are stored in similar way as JPEG compression and P-frames are computed as predicted motion vectors and the prediction error between the frame to be compressed and the motion predicted version of the frame.

My question is: how B-frames are computed and stored? Every search that I've made describes it as "motion vectors and transform coefficients" and they are "predicted or interpolated from an earlier and/or later frame". I know that two motion vectors are computed for each macroblock, but how I use them? Are these vectors coded and transmitted as in the case of a P-frame? The prediction error of the b-frame is between which estimatives?

Hanz0
  • 59
  • 7
  • possible duplicate of [Are B, P frames results of the motion estimation/compression?](http://stackoverflow.com/questions/23595408/are-b-p-frames-results-of-the-motion-estimation-compression) – n0p Mar 25 '15 at 09:16

1 Answers1

3

The B-frames are both, forward predicted and backward predicted from the last/next I-frame or P-frame

The following picture describe how it going to predict. enter image description here

nifCody
  • 2,394
  • 3
  • 34
  • 54