0

As the title, currently most of algorithms for video compression choose to use 2D-DCT and motion compensation (and some other techniques), which forms a standard.

Here is the question: Why not use 3D-DCT for video compression? And what is the shortage for 3D-DCT comparing to 2D-DCT for compression?

Many thanks ~

sissuire
  • 1
  • 1

2 Answers2

0
  1. frames must be processed in a group instead of a single frame
  2. the computation of the 3D DCT is more complex than 2D DCT
Raymond K W Chan
  • 1
0

In versatile video coding, the partitioning to apply transformation is applied adaptively for one coding block (which is a rectangular block in a frame which is 2D). In order to be able to use 3D DCT we would need cuboid. Coding block partitioning is by default not same in consecutive frames, hence not straight forward to find a proper cuboid to apply 3D DCT. One can try to see if applying similar partitioning for residual coding in consecutive frames (consecutive in terms of may be processing or video input), would be beneficial but it's unlikely to work. Because the statistics of the residual signal over frames would be very different due to combination of inter and intra prediction.

pd_25
  • 1