International organisation for standardisation organisation internationale de normalisation



Yüklə 2,78 Mb.
səhifə41/67
tarix02.01.2022
ölçüsü2,78 Mb.
#20862
1   ...   37   38   39   40   41   42   43   44   ...   67

4Explorations – 3D Video

A goal of the 3D video exploration activity is to enable the generation of interpolated views from available source views of multiview camera configurations. The target application is mostly seen as being for upcoming generations of various (auto-) stereoscopic displays, either requiring multiple views internally or providing means for baseline adjustment. In the new envisioned scheme, only a low number (1-3) of video sequences would be transmitted, and the rendering of additional views would be enabled by associated depth information.


In the exploration experiments preceding the London meeting, further progress was made on improving (automatic) depth estimation and view rendering / interpolation. Two approaches were investigated, namely layered depth video (LDV, including background video plus depth) and stereo/multiview plus depth (MVD), both of which seem to be conceptually working. For the upcoming CfP, it was decided to concentrate from now on on one technology for producing anchors, namely MVD using MVC coding for both video and depth in parallel.
For the first time, reasonable results were achieved in work on "Exploration Experiment number 4" (EE4), which is the coding experiment work, by performing the second step in anchor preparation for a CfP, which is coding experiments, using AVC/MVC for both video and depth map coding. However, due to the high amount of data that were provided (combining 5 rate points each for depth and video, which amounts to 25 combinations), it was not possible to select best combinations by viewing experiments. However, it was possible to identify useful ranges of rates for each sequence, by comparing the subjective behaviour at lowest and highest rate points. This is reflected in the new EE definition. To speed up the selection of optimum rate allocation for the video and depth components, participants of EE4 were requested to perform a preview at their home lab, and make a proposal about depth/video rate combination case by case.
Further, the document “Applications and Requirements of 3D Video Coding” was slightly updated to better distinguish between representation and input/output formats (N 10857).
Regarding the application cases of 3D Video coding, it is currently unclear whether the two envisaged application scenarios – 1. Enhanced Stereo (baseline adaptation), and 2. future enhanced view-range displays (multiview, autostereoscopic) – would require identical timelines, an also whether their technical requirements are identical. National bodies are kindly requested to provide more information on this, and also the Exploration Experiments (EEs) are now defined around the 2 scenarios to become able of better judging the maturity of available technical solutions.
The planned exploration experiments are as follows, targeting the preparation of material for the CfP based on MVD technology (N 10720):

  • EE1: Depth map generation

    • Quality for 2-view / 2x baseline distance case has become good

    • Generation of additional depth maps for 3-view / 4x baseline case

    • Further improvement of interpolation methods

  • EE4: Coding experiments

    • Find best data rate combination points (video, depth) for possible CfP

    • Data rates defined per sequence; are sequences sufficiently "difficult"?

In particular, multiview sequences captured with moving cameras are needed, and in general it appears that some of the sequences available have too large static background parts.


Documents reviewed (also see AHG meeting report)


m16576

Results of Exploration Experiments in 3D Video Coding

Yongzhe Wang
Hunn Rhee
Philipp Merkle
Karsten Müller

m16582

Improved 1D Mode in VSRS 3.1

Yin Zhao
Lu Yu
Dong Tian

m16583

Enhanced LDV Background Information Generation

Yin Zhao
Lu Yu
Lianhuan Xiong

m16584

Peak Signal-to-Perceptible-Noise Ratio Tool: PSPNR 1.0

Yin Zhao
Lu Yu

m16585

3DV EE4 Report on Champagne tower sequences (LDV data)

Yin Zhao
Lu Yu

m16590

Report of 3DV Exploration Experiment with Champagne Tower

Taka Senoh
Kenji Yamamoto
Ryutaro Oi
Tomoyuki Mishina
Taiichiro Kurita

m16591

3DV/FTV EE report on "Lovebird1"

Gi-Mun Um
Gun Bang
Namho Hur
Jinwoong Kim

m16599

3DV EE4 on BeerGarden sequence

Patrick Lopez
Dong Tian
Po-Lin Lai

m16601

3DV EE1 & EE4 Results on Newspaper sequence

Seok Lee
Jaejoon Lee
Ilsoon Lim
Du-Sik Park

m16605

Depth Estimation Reference Software (DERS) 4.0

Masayuki Tanimoto
Toshiaki Fujii
Mehrdad Panahpour Tehrani
Menno Wildeboer

m16606

3DV/FTV EE1 results of "Champagne tower"

Masayuki Tanimoto
Toshiaki Fujii
Mehrdad Panahpour Tehrani
Menno Wildeboer

m16607

Error Cancellation in Free-viewpoint Image Generation for FTV

Masayuki Tanimoto
Toshiaki Fujii
Mehrdad Panahpour Tehrani
Menno Wildeboer
Hisayoshi Furihata

m16615

3DV/FTV EE1 Report on Pantomime sequence

Shinya Shimizu
Hideaki Kimata

m16642

3DV EE results on BookArrival test sequence

Jaewon Sung
Younghee Choi
Byeong-Moon Jeon

m16643

A View Synthesis Tool for 3D video

Jaewon Sung
Byeong-Moon Jeon

m16646

Philips contribution on extended 3DV LDV Test Material

fons.bruls@philips.com
rene.klein.gunnewiek@philips.com
patrick.vandewalle@philips.com

m16647

Extended 3DV LDV Test Material for Newspaper

fons bruls
rene klein gunnewiek
patrick van de walle
Yin Zhao

m16670

Results of EE1 on 'Pantomime? Sequence

Yo-Sung Ho



m16671

Results of EE4 on ?Newspaper? Sequence

Yo-Sung Ho



m16673

Bilateral Filtering of Depth Map for 3D Video Coding

Yo-Sung Ho



m16674

Temporally Enhanced 3D Test Sequence: Develiry

Yo-Sung Ho



m16675

3DTV Exploration Experiments on Pantomime data set

Ivana Radulovic
Andrey Norkin
Per Fröjdh

m16709

3DV EE4 on Mobile

Sehoon Yea
Anthony Vetro

m16718

3DV EE4 on Champagne_tower

Carmen CHENG
Yan HUO
Yu LIU

m16756

Results of Exploration Experiments in 3D Video Coding, described in w10552, for ?Newspaper? sequence

Olgierd Stankiewicz
Krzysztof Wegner

m16757

An enhancement of Depth Estimation Reference Software with use of soft segmentation

Krzysztof Wegner
Olgierd Stankiewicz

m16758

Influence of views and depth compression onto quality of synthesized views

Krzysztof Klimaszewski
Krzysztof Wegner
Marek Domanski



Yüklə 2,78 Mb.

Dostları ilə paylaş:
1   ...   37   38   39   40   41   42   43   44   ...   67




Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur ©muhaz.org 2024
rəhbərliyinə müraciət

gir | qeydiyyatdan keç
    Ana səhifə


yükləyin