Improved perceptual video encryption using multiple 8×8 transforms in MPEG-4

Perceptual video encryption by making use of multiple transforms alternatively (according to the random encryption key) has been demonstrated to be quite successful in H.264 video coding. An important reason is that both intra and inter frames in H.264 need go through a prediction (intra-prediction or motion-compensation) so that different transforms (from DCT) always work on some residual signals, leading to nearly the same coding efficiency (as DCT). This technique can be applied to all inter-frames in MPEG-4, but not to any intra-frame because the intra-predication is performed in the transform domain in MPEG-4. In this paper, we present a solution to this problem: mapping the intra-predicated transform coefficients back to the pixel domain. Based on the equivalent intra-prediction in the pixel domain, we can then apply our multiple transform based encryption scheme on each intra-frame. Experiment results show that encrypting each intra-frame in MPEG-4 provides a further quality-degradation on the encrypted video, thus making this multiple transform based scheme applicable in more applications.