EE 5359-MULTIMEDIA PROCESSING

EE 5359-MULTIMEDIA PROCESSING

3D EXTENSION of HEVC: Multi-View plus Depth

Parashar Nayana KarunakarStudent Id: 1000833406Department of Electrical Engineering

HEVC & MVD- BRIEF OVERVIEW

• High-Efficiency Video Coding (HEVC) is the newest video coding standard of the ITU-T Video Coding Experts Group (VCEG) and the ISO/IEC Moving Picture Experts Group (MPEG).

• The Joint Collaborative Team on 3D Video Coding Extension Development was created to develop 3D video coding technology more advanced than the current multiview video coding (MVC) features of H.264. The standards for which these 3D video coding extension technologies will provide such enhanced capabilities may include H.262, H.264 and the High Efficiency Video Coding (HEVC) .

HEVC – ENCODER BLOCK DIAGRAM

Encoder block diagram H.264 [18]

Fig 1. – Typical HEVC Encoder[1]

What is 3D/Multi-View/Stereo Video• Multiple Cameras views of the same scene are captured – Multiple View Video(MVV)

• Efficient Compression techniques are essential as MVV contains vast amount of data both during storage and transmission.

• Inter-view statistical dependencies are exploited for combined temporal/inter-view prediction.

• When color video and an associated per sample depth map are considered, we get Multi-view video plus depth representation.

Fig 2. – Test Sequence –Balloons with depth map [3]

Fig 3. – Overview of the system structure and data format for the transmission of3D video [12]

MVC FOR MULTI-VIEW PLUS DEPTH• Encoding and decoding of each view of a multi-view data set separately ,referred to as simulcast coding , can be done with any video codec including H.264/AVC and HEVC.

• This would be simple but inefficient as inter-view statistical dependencies are not exploited.

• In order to exploit all the statistical dependencies within multi-view dataset, inter-view prediction has to be combined with temporal prediction.

• As seen in fig 3b. , In MVC, one of the views is conventionally coded in conformance to the HEVC codec. For coding the other views, in addition to previously coded pictures of the same view already coded co-located pictures of other views can also be used as reference pictures.

• In Multi –view video plus depth (MVD) format, only a few views are actually coded. Based on the transmitted videos and depth maps, additional views can be rendered..

Fig 4a. – Simulcast coding structure with hierarchical B pictures for temporal prediction(black arrows)[2]

Fig 4b. – Multi-view coding structure with hierarchical B pictures for both temporal (black arrows) and inter-view prediction(red arrows) [2]

BASIC 3D VIDEO CODEC STRUCTURE

Fig 5. – Block Diagram of a 3D Video Codec[4]

MVD CODEC- WORKING• The basic structure of the 3D video codec is shown in the block diagram of Figure 5. In principle, each component signal is coded using an HEVC-based codec. The resulting bit stream packets, or more accurately, the resulting Network Abstraction Layer (NAL) units, are multiplexed to form the 3D video bit stream.

• The base or independent view is coded using an unmodified HEVC codec. The base view sub-stream can be directly decoded using the conventional HEVC decoder.

• For coding the dependent views and the depth data, modified HEVC codec are used, which are extended by including additional coding tools and inter-component prediction techniques that employ already coded data inside the same access unit as indicated by the red arrows in Figure 5.

• For enabling an optional discarding of depth data from the bit stream, e.g., for supporting the decoding of a stereo video suitable for conventional stereo displays, the inter-component prediction can be configured in a way that video pictures can be decoded independently of the depth data..

MVD- CODING ALGORITHM• The video pictures and, when present, the depth maps are coded access unit by access unit, as it is illustrated in Figure 6.• An access unit includes all video pictures and depth maps that correspond to the same time instant. NAL units containing camera parameters may be additionally associated with an access unit.• The video pictures and depth maps corresponding to a particular camera position are indicated by a view identifier (viewId). All video pictures and depth maps that belong to the same camera position are associated with the same value of viewId.• Inside an access unit, the video picture and, when present, the associated depth map with viewId equal to 0 are coded first, followed by the video picture and depth map with viewId equal to 1, etc.• For ordering the reconstructed video pictures and depth map after decoding, each value of viewId is associated with another identifier called view order index (VOI). The view order index is a signed integer values, which specifies the ordering of the coded views from left to right.

Fig 6. - Access units structure and coding order of view components[12]

COMPARSION – MVD AND HEVC CODEC• CODING OF DEPENDENT VIEWS -- Additional tools have been integrated into the HEVC codec, which employ already coded data in other views for efficiently representing a dependent view. These tools include - Disparity-compensated prediction, View synthesis based inter-view prediction, Post processing in-loop filtering, Inter-view motion prediction, Depth-based motion parameter prediction, Inter-view residual prediction, Adjustment of QP of texture based on depth data.

• CODING OF DEPTH MAPS – There are certain additional tools and also some tools are removed for coding of Depth maps. Some of the differences are -- Depth Maps are coded in 4:0:0 format, Non-linear depth representation is used, Z-near Z-far compensated weighted prediction, Modified motion compensation and motion vector coding ( No interpolation is used i.e. for depth maps, the inter-picture prediction is always performed with full-sample accuracy. Disabling of in-loop filtering ( deblocking filter and SAO), Depth modeling modes ( Four new Intra-prediction modes are used), Motion parameter inheritance.

THE PROPOSAL

• The aim of the project is to import the MVD coding tools into the HM 9.1.

• The project will be carried out in the following steps:

Learn about the tools that are different to MVD when compared to standard HEVC codec.

Study and compare the MVD extensions of HEVC carried out on HM 5.1by Fraunhoffer HHI and Qualcomm ( the two most recent ones).Import the MVD to HM 9.1 based on the observations from the first two

stages.

Present the results and report the changes in bitrates and PSNR between the new imported software and the previously studied

INTERIM RESULTS• The sequence balloons (1024 x 768) has been used. The tested configuration is the balloons sequence with three camera view including depth with QP of 30.

• The encoding is done using both HTM 5.1 and HTM 6.0 to compare them on this system.

• The results are as shown in the next few slides,

INTERIM RESULTS – HTM 5.1 V/S HTM 6.0

Encoding Time

0500

1000150020002500300035004000

HTM 5.1HTM 6.0

INTERIM RESULTS – HTM 5.1 V/S HTM 6.0 BIT-RATE

Video 0 Depth 0 Video 1 Depth1 Video 2 Depth20

200400600800

10001200140016001800

HTM 5.1HTM 6.0

INTERIM RESULTS – HTM 5.1 V/S HTM 6.0 Y-PSNR

Video 0 Depth 0 Video 1 Depth1 Video 2 Depth205

1015202530354045

HTM 5.1HTM 6.0

VIDEO USED – BEFORE AND AFTER COMPRESSION – HTM 5.1

Artifacts

VIDEO USED – BEFORE AND AFTER COMPRESSION – HTM 6.0

ArtifactsArtifacts

OTHER WORK DONE

• Importing the 3D compression features into the HM 9.2 has been started.

• Unfortunately, results can only be shown once complete porting of the encoder is complete.• HTM 6.0 is referred when porting the 3D compression techniques to 9.2

LIST OF ACRONYMSAVC: Advanced video codingITU: International telecommunication unionMPEG: Moving picture experts groupMVC: Multi-View CodingMVD: Multi-View plus DepthMVV: Multi-View VideoSAO: Sample Adaptive OffsetVCEG: Video coding experts groupVOI: View Order Index

REFERENCES[1] Overview of the High Efficiency Video Coding (HEVC) Standard Gary J. Sullivan, Fellow, IEEE, Jens-Rainer Ohm, Member, IEEE, Woo-Jin Han, Member, IEEE, and Thomas Wiegand, Fellow, IEEE.

[2] P. Merkle, A Smolic, K. Müller, and T. Wiegand, “Multi-View video plus depth data representation and coding”. PCS 2007

[3] http://www.tanimoto.nuee.nagoya-u.ac.jp/~fukushima/mpegftv/

[4] H. Schwarz, C. Bartnik, S. Bosse, H. Brust, T. Hinz, H. Lakshman, D. Marpe, P. Merkle, K. Müller, H. Rhee, G. Tech, M. Winken, and T. Wiegand, "3D Video Coding Using Advanced Prediction, Depth Modeling, and Encoder Control Methods", Picture Coding Symposium, May 2012.

[5] G. Tech, H. Schwarz, K. Müller, and T. Wiegand, "Effects of synthesized View Distortion based 3D Video Coding on the Quality of interpolated and extrapolated Views", IEEE Intl. Conf. on Multimedia and Exposition, July 2012.

[6] P. Merkle, C. Bartnik, K. Müller, D. Marpe, and T. Wiegand, "3D Video: Depth Coding Based on Inter-component Prediction of Block Partitions", Picture Coding Symposium, May 2012.

[7] H. Schwarz and T. Wiegand, "Inter-View Prediction of Motion Data in Multiview Video Coding", Picture Coding Symposium, May 2012.

[8] G. Tech, H. Schwarz, K. Müller, and T. Wiegand, "3D Video Coding using the Synthesized View Distortion Change", Picture Coding Symposium, May 2012.

[9] M. Winken, H. Schwarz, and T. Wiegand, "Motion Vector Inheritance for High Efficiency 3D Video plus Depth Coding," Picture Coding Symposium, May 2012.

[10] S. Bosse, H. Schwarz, T. Hinz, T. Wiegand, "Encoder Control for Renderable Regions in High Efficiency Multiview Video Plus Depth Coding", Picture Coding Symposium, May 2012.

[11] hevc.info

[12] “Test Model under Consideration for HEVC based 3D video coding”, ISO/IEC JTC1/SC29/WG11 MPEG2011/N12559February 2012, San Jose, CA, USA