International organisation for standardisation organisation internationale de normalisation

5.2.43D-HEVC-CE4: Transform Coding for Depth

m23909	Gun Bang	3D-HEVC-CE4 Summary Report
m23621	Gun Bang, Junghak Nam, Hyomin Choi, Sunmi Yoo, Woong Lim, Won-Sik Cheong, Namho Hur, Donggyu Sim	3D-HEVC-CE4 results on adaptive quantization for depth map
m23653	Byeongdoo Choi, Chanyul Kim, Jeonghoon Park	Cross check of 3D-HEVC-CE4 results on ETRI's proposal

Summary report (3 tools considered):

• 
  A 64x64 transform for depth as proposed in m22697 by Poznan University (not tested for this meeting).
  
• 
  Quantization in the sample somain (without transform) in m22625 as proposed by ETRI&KWU
  
• 
  Rotational transform for depth coding in m22634 as proposed by Samsung (the rotational transform for depth coding could not be provided, so the proposal was dropped)
  

m23621 contained ETRI and KWU’s proposal on adaptive quantization for depth maps, using a lower QP for the spatial quantization than the QP for the frequency domain, yielding 2.9% and 1.7% bit rate reductions for depth maps in 2-views and 3-views cases, with the same encoder/decoder complexity. No gains were reported in the synthesized view domain.

It was suggested to continue the CE with an emphasis on the 64x64 transform for depth until the next meeting, and conducting further studies on ETRI and KWU’s proposal.

Conclusions:

• 
  Discontinue CE.
  
• 
  Encourage proposal on the benefits of the 64x64 transform for depth.