Home

  ‰p•¶ì¬ƒeƒ“ƒvƒŒ[ƒg

The 16*16 reconstructed prediction error blocks are obtained by up-sampling the 8*8 reduced-resolution reconstructed prediction error blocks. (...) Up-Sampling of the reduced-resolution reconstructed prediction error The 16*16 reconstructed prediction error block is obtained by up-sampling the 8*8 reduced-resolution reconstructed prediction error block. (...) FIGURE 7/Annex Q Positioning of samples in 8*8 reduced-resolution reconstructed prediction error block and 16*16 reconstructed prediction error block Q.6.1.

Score: 679093.1 - https://www.itu.int/wftp3/av-a...deo-site/9702_Nic/lbc97021.doc

  WMO Polar Prediction Project: Year Of Polar Prediction | World Meteorological Organization

WMO Polar Prediction Project: Year Of Polar Prediction | World Meteorological Organization Skip to main content World Meteorological Organization Weather · Climate · Water Toggle navigation العربية 简体中文 English Français Русский Español Go Our mandate What we do Weather How we do it Public-Private Engagement (PPE) ocp innovation webinar Space borne Precipitation Measurements and Application WMO and HMEI Information Day WMO Innovation Seminar - Microsoft and the UN Sustainable Development Goals Climate Focus areas Water Programmes Projects Resources Bulletin MeteoWorld Library Gender Equality Education and training Standards and Recommended Practices The WMO Building / Conference Centre Language resources World Meteorological Day United in Science Coronavirus (COVID-19) Media Events About us Who we are Vision, Mission, Strategic Planning Members Governance The Secretariat Employment Awards Procurement Finance and Accountability Related links FAQs Contact us Community Platform Reform Search form Search Home Evolution%20des%20temp%C3%A9ratures%20moyennes%20minimales%20et%20maximales%20quotidiennes%20en%20France%20-%20janvier%202015%20-%20Taille%20r%C3%A9duite.png WMO Polar Prediction Project: Year Of Polar Prediction WMO Polar Prediction Project: Year Of Polar Prediction Video of WMO Polar Prediction Project: Year Of Polar Prediction Follow WMO Discover Events News Bookstore Projects Bulletin MeteoWorld Learn Visit the Library Youth Corner WMO Governance Partnerships Contact us Procurement Privacy policy Report fraud, corruption or abuse About us Disclaimer Copyright Sitemap © 2022 World Meteorological Organization (WMO)

Score: 678928.12 - https://public.wmo.int/en/file...oject-year-of-polar-prediction

  Multi-hypothesis prediction

In the proposed method, the merging process is combined with motion hypothesis prediction. When enabled, the current partition is predicted with the weighted sum of the two candidate partitions. (...) In this contribution, the merging process is extended by using motion-hypothesis inter prediction. It is known that the accuracy of motion compensated prediction can be increased with multi-hypothesis prediction (motion-compensated prediction using multiple motion parameters) [3]. (...) Figure.2 Motion hypothesis prediction In addition, the current partition needs to have motion parameters because it will be used for prediction of right or below neighbor partition in the future.

Score: 678688.67 - https://www.itu.int/wftp3/av-a..._07_B_Geneva/JCTVC-B023_r2.doc

  ITU - Telecommunications Standardization Sector

This means that if negative rounding is used for forward prediction and positive rounding is used for backward prediction in motion compensation for frame 3, the same interpolated reconstructed images of frames 2 and 4 can be used multiple times for the prediction of P-pictures and B-Pictures [3]. (...) The RTYPE bit for B-pictures indicate the rounding method for forward prediction. For backward prediction, the opposite method from forward prediction is used. (...) For backward prediction, the method opposite from forward prediction is applied (can be achieved by syntax (3), (4), or (5) shown in section 3).

Score: 678688.67 - https://www.itu.int/wftp3/av-a...video-site/9709_Sun/q15b39.doc

  Nokia Standard Document Template

In [1] the quantization value for the predicted block was kept the same as the quantization value for the prediction error. (...) In this document we further improve the coding efficiency of SP-frames by using a separate quantization value for the predicted frame than the prediction error coefficients. (...) Conclusion We have presented a new SP-picture encoding/decoding technique which uses different quantization values for the predicted block and the prediction error coefficients.

Score: 678688.67 - https://www.itu.int/wftp3/av-a...deo-site/0104_Aus/VCEG-M73.doc

  Joint Video Team (MPEG+ITU) Document

The solid-line arrows show conventional predictions, namely, prediction from I- or P-pictures. The dotted line arrows show the proposed additional prediction, namely, prediction from B-pictures. If the enhanced bi-directional prediction mode is employed, several items need to be studied and modified. (...) B1 P2 B3 B4 P5 B6 B7 P8 B9 Conventional prediction Additional prediction Figure 1: Additional new prediction method for B-pictures. 2.2.

Score: 678445.4 - https://www.itu.int/wftp3/av-a...deo-site/0201_Gen/JVT-B057.doc

  Joint Video Team (MPEG+ITU) Document

The solid-line arrows show conventional predictions, namely, prediction from I- or P-pictures. The dotted line arrows show the proposed additional prediction, namely, prediction from B-pictures. If the enhanced bi-directional prediction mode is employed, several items need to be studied and modified. (...) B1 P2 B3 B4 P5 B6 B7 P8 B9 Conventional prediction Additional prediction Figure 1: Additional new prediction method for B-pictures. 2.2.

Score: 678445.4 - https://www.itu.int/wftp3/av-a...te/2002_01_Geneva/JVT-B057.doc

  Establishment of a Climate Prediction Analysis System in Uganda | World Meteorological Organization

Establishment of a Climate Prediction Analysis System in Uganda | World Meteorological Organization Skip to main content World Meteorological Organization Weather · Climate · Water Toggle navigation English Go Our mandate What we do Weather How we do it Public-Private Engagement (PPE) ocp innovation webinar Space borne Precipitation Measurements and Application WMO and HMEI Information Day WMO Innovation Seminar - Microsoft and the UN Sustainable Development Goals Climate Focus areas Water Programmes Projects Resources Bulletin MeteoWorld Library Gender Equality Education and training Standards and Recommended Practices The WMO Building / Conference Centre Language resources World Meteorological Day United in Science Coronavirus (COVID-19) Media Events About us Who we are Vision, Mission, Strategic Planning Members Governance The Secretariat Employment Awards Procurement Finance and Accountability Related links FAQs Contact us Community Platform Reform Search form Search Home المشاريع Establishment of a Climate Prediction Analysis System in Uganda Establishment of a Climate Prediction Analysis System in Uganda 9303840310_1cff89316f_o.jpg Establishment of a Climate Prediction Analysis System in Uganda Project Type National Projects WMO Strategic Priority: Global Framework for Climate Services Capacity Development Status: Ongoing Funding:   150,000.00 CHF Implementation in: Uganda Regions: Region I: Africa Donors: Korea Meteorological Administration (KMA) Project Partners: IGAD Climate Prediction and Applications Centre (ICPAC) Contact: Climate prediction and adaptation requires scientifically sound weather and climate information and capacity to develop and use such information for decision-making at national level. The Korea Meteorological Administration (KMA) is funding several projects to enhance such capacity in the countries in Eastern Africa that are members of the Intergovernmental Authority on Development (IGAD) Climate Prediction Applications Center (ICPAC). Climate Prediction Analysis Systems will be set up as part of these projects in Rwanda, Uganda, Burundi and Djibouti as initial pilot countries. Donors Project Partners You may also find this interesting WMO Project Establishment of a Climate Prediction Analysis System in Rwanda Follow WMO Discover Events News Bookstore Projects Bulletin MeteoWorld Learn Visit the Library Youth Corner WMO Governance Partnerships Contact us Procurement Privacy policy Report fraud, corruption or abuse About us Disclaimer Copyright Sitemap © 2022 World Meteorological Organization (WMO)

Score: 678262.74 - https://public.wmo.int/en/proj...diction-analysis-system-uganda

  2010041321064795941_Microsoft PowerPoint - 100410_JCTVC-A110.pptx

2010041321064795941_Microsoft PowerPoint - 100410_JCTVC-A110.pptx LG’s proposal for HVCLG’s proposal for HVC JCTVC-A110 LG Electronics ByeongMoon Jeon, SeungWook Park, JungSun Kim, JoonYoung Park Introduction v Based on AVC/H.264, the following new features are added ß Large macroblock structure (64x64 to 8x8 coding blocks) ß IIMM (Inter-Intra Mixed Mode) ß Skip mode with variable block size ß SPMV (Scaled Predicted Motion Vector) ß Template-based IC(Illumination Compensation) ß Modified MVC (Motion Vector Competition) ß SIFO of KTA ß New intra prediction types (I_32x32, I_mixed, ..,etc) ß New chroma intra prediction mode (chroma estimation mode) ß Border handling scheme ß ADF (Adaptive Deblocking Filter) ß QALF of KTA ß MDDT with modified kernels ß Adaptive scan ordering ß AWR (Adaptive Warped Reference) ß PAIF (Parametric Adaptive Interpolation Filter) ß MVC with B skip/direct v Proposed model has Substantially increased compression capability relative to AVC/H.264 Macroblock Structure v Macroblock structure for inter prediction ß Macroblock unit size is 32x32 ß partitioned into several shapes of motion-compensated blocks ß Can be clustered up to 64x64 32x3232x6464x3264x64 16x1616x3232x1632x32 8x88x1616x816x16 Macroblock Structure (Cont’d) v Skip mode with variable block size ß Applies to the following blocks partitions ÿ 64x64, 64x32, 32x64, 32x32 ‡ efficient for high resolution ÿ 32x16, 16x32, 16x16 ‡ efficient for low resolution ß Increases the portion of skip mode ÿ Increases coding efficiency 32x3232x6464x3264x64 SKIP P64x32 SKIP SKIPSKIP P32x32 P32x16 P32x16 SKIP SKIP P16x16 P16x16 SKIP SKIP ß A macroblock (32x32) has both inter and intra submacroblocks ß 1 bit flag in each submacroblock (16x16) to signal inter or intra ß Same structure with AVC/H.264 ‡ efficient for low resolution v IIMM (Inter-Intra Mixed Mode) Macroblock Structure (Cont’d) I_4x4 I_16x16Inter 8x16 Inter 8x8 32 32 v SMVP (Scaled motion vector predictor) ß The motion vectors of the neighboring blocks is scaled according to the temporal distance between current and reference pictures Scaled mvLXN = (td / tb) * mvLXN tb td mvL0N Scaled mvL0N neighboring partition current partition ref ref current Inter prediction ß To compensate the illumination change between pictures ß Offset value representing the illumination change between current and predicted block is derived at decoder ÿ Offset is set to the difference value between the DC values of Tcur and Tpred ß Offset value is added to the predicted block to compensate illumination change Inter prediction (Cont’d) refPicL0 currPic predBlk curBlk mvL0 Tpred Tcur ＋ offset ＋ resBlk predBlk’ recBlk＋ _ v Template-based IC (Illumination Compensation) Inter prediction (Cont’d) v Modified MVC (Motion Vector Competition) ß MVC of KTA is modified to be compatible with the new tools ÿ Combined with SPMV(Scaled predicted motion vector) v SIFO(Switched Interpolation Filter Offset) ß The single pass SIFO of KTA is employed Intra-frame prediction v Intra prediction types ß A macroblock has one of two intra macroblock modes ÿ I_32x32, I_mixed ß Each submacroblock has one of three submacroblock modes ÿ I_4x4, I_8x8, I_16x16 I_mixedI_32x32 I_8x8I_4x4 I_16x16 v I_32x32 ß Four prediction modes (vertical, horizontal, DC, plane) of AVC/H.264 is used for 32x32 block Intra-frame prediction (cont’d) Intra32x32PredMode Name of Intra32x32PredMode 0 Intra_32x32_Vertical 1 Intra_32x32_Horizontal 2 Intra_32x32_DC 3 Intra_32x32_Plane ß Each submacroblock can be coded one of three prediction modes, ÿ I_4x4, I_8x8, I_16x16 ÿ Each mode has one of 9 prediction mode of AVC/H.264 ß Two flags (transform_16x16_flag , transform_8x8_flag) signal prediction type and transform size of a submacroblock Intra-frame prediction (Cont’d) I_mixed I_8x8I_4x4 I_16x16 transform_16x16_flag transform_8x8_flagI_16x16 16x16 transform (DCT or MDDT) I_8x8 8x8 transform (DCT or MDDT) I_4x4 4x4 transform (DCT or MDDT) 1 0 1 0 v I_mixed v Chroma intra prediction ß Prediction unit size is 8x8 ß New chroma prediction mode (chorma estimation mode) is added to AVC/H.264 modes (DC, vertical, horizontal, plane) ß Predicted chroma samples are derived from the reconstructed luma samples based on linear model Intra-frame prediction (Cont’d) )'(*)'( )','( )','( LC LL CL PmeanPmean PPR PPR αβ α −= = predc[ x, y ] = α * PL[ 2*x, 2*y ] + β predC[ x, y ] : chroma pixels to be predicted PL[ 2*x, 2*y ] : subsampled luma pixels ‡ gray pixels in luma block P’L : pixels subsampled from neighboring pixels of luma block ‡ black pixels in luma block P’C : neighboring pixels of chroma block ‡ black pixels in chroma block Luma block Chroma block v PROPOSED METHOD ß When a part of macroblock (32x32) lies within the padded region, the macroblock is partitioned into four 16x16 submacroblocks ß Submacroblocks within the padded region are derived at decoder without any overhead v P / B Picture ÿ 16x16 SKIP v I Picture ÿ I_16x16 ÿ prediction mode depends on its position • right border : horizontal mode • bottom border : vertical mode Border handling scheme normal coding normal coding 416 24 0 16 16x16 SKIP 16x16 SKIP normal coding P / B Picture I_16x16 25 6 normal coding normal coding I_16x16 normal coding I Picture I_mixed I_16x16 v Transforms ß DCT : 4x4, 8x8, 16x8, 8x16, 16x16 ß MDDT : New MDDT kernels Transform / Scan Order Luma Chroma Block size Transform Transform Inter 64x64, 64x32, 32x64, 32x32, 32x16, 16x32, 16x16 ß 4x4, 8x8, 16x16 ß 4x4 ß 2x2 or 4x4 DC Hadamard depending on skip status 16x8 ß 4x4, 8x8, 16x8 8x16 ß 4x4, 8x8, 8x16 8x8 ß 4x4, 8x8 Intra I_32x32 ß DC pred mode : 16x16 DCT ß Other modes : 16x16 MDDT ß 4x4 ß 4x4 DC HadamardI_mixed (I_16x16, I_8x8, I_4x4) Same transform size as block size v Scan order ß Adaptive scan order with new initial values In-loop filters / Entropy Coding v In-loop Filters ß Adaptive Deblocking Filter (ADF) ÿ Adaptive deblocking filter based on ‘Wiener filtering’ scheme ÿ Filter coefficients are calculated by resolving MMSE between the original and the reconstructed pictures. ÿ ADF or AVC/H.264 deblocking filter is applied adaptively in each slice ß Quad-tree based Adaptive Loop Filter (QALF) ß QALF of KTA is employed v Entropy Coding ß AVC/H.264 CABAC is employed Other tools u Tools being developed but not included in the submitted model u By adopting these tools, further improvements can be expected v AWR (Adaptive Warped Reference) ß Generates a new warped reference picture considering complex motions such as scaling, rotation, sheering, and so on. ß Modeling of a parametric warping function using KLT(Kanade-Lucas-Tomasi) feature tracker ß Refer to JCTVC-021 v PAIF (Parametric Adaptive Interpolation Filter) ß Transmits just 5 parameters instead of individual filter coefficients to represent interpolation filter ß Less bits for representing filters and closer to optimal filter than existing AIFs ß Refer to JCTVC-021 v MVC (Motion Vector Competition) with B skip/direct ß MVC with B skip/direct is utilized to maximize benefits of MVC scheme v Chroma estimation mode with phase shift ß The new chroma prediction mode (chroma estimation mode) is modified considering phase difference between luma and chroma samples Coding structure Constraint set 1 v Hierarchical B coding / GOP8/ IDR picture every 1 sec v Same configuration of alpha anchor except the following changes ß Number of reference frames for P pictures=5 ß Number of reference frames for B pictures=4 (2 reference pictures allowed for each list) ß Weighted prediction disabled ß 16x16, 16x8, 8x16, 8x8, 4x4 transforms enabled Group of picture (GOP) I PB1B2 B2B3 B3 B3 B3 PB1B2 B2B3 B3 B3 B3 I I PPP PP P P P Constraint set 2 v I-P-P-P coding / IDR only at first picture v Same configuration of gamma anchor except the following changes ß Number of reference frames for P pictures=5 ß CABAC enabled ß 16x16, 16x8, 8x16, 8x8, 4x4 transforms enabled 31 32 33 34 35 36 0 2000 4000 6000 8000 10000 12000 Y PS N R (d B ) Rate (Kbps) Class B-BQTerrace LG JM 30 32 34 36 38 40 0 300 600 900 1200 1500 1800 Y PS N R (d B ) Rate (Kbps) Class D-RaceHorses LG JM RD performance v Constraint Set 1 (alpha) Class BD psnr BD rate A 1.11 -23.94 B 1.01 -30.57 C 1.26 -26.85 D 0.92 -19.85 Avg. 1.06 -25.83 MaxMax MinMin BD-PSNR BD-Rate 0.71 -36.15 BD-PSNR BD-Rate 0.62 -11.76 29 30 31 32 33 34 35 36 0 2000 4000 6000 8000 10000 12000 Y PS N R (d B ) Rate (Kbps) Class B-BQTerrace LG JM 28 30 32 34 36 38 40 42 0 300 600 900 1200 1500 1800 Y PS N R (d B ) Rate (Kbps) Class D-RaceHorses LG JM RD performance v Constraint Set 2 (gamma) Class BD psnr BD rate B 1.85 -44.42 C 1.60 -33.65 D 1.15 -25.11 E 2.31 -45.01 Avg. 1.70 -37.01 MaxMax MinMin BD-PSNR BD-Rate 1.68 -52.15 BD-PSNR BD-Rate 0.79 -14.99 Subject quality comparison CS1 - Class C PartyScene @R2 (512kbps) v An artifact (propagation & blurring of chroma components) are visible especially when coded at low bit rates v The artifact is removed by considering chroma components in the R-D calculation process at encoder side Alpha anchor Proposed model Subject quality comparison (Cont’d) CS1 - Class C BasketballDrill @R2 (512kbps) Alpha anchor Proposed model Subject quality comparison (Cont’d) CS1 - Class C BasketballDrill @R2 (512kbps) Alpha anchor Proposed model Software implementation v On top of JM 11.0, new tools are implemented v C programming language v Complied using Microsoft Visual Studio 2008 under Microsoft Widows XP 32bit edition platform v No code-level optimization for complexity Complexity v Time Measurement ß Consumed time for encoding & decoding ß Platform ÿ 32bit executables are used in encoder and decoder. ÿ CPU : Intel Core i7 950 @ 3.07Ghz (quad core) ÿ Memory : 12 GB ß Two target rates are tested due to time limitation ÿ Rate2 (middle rate) / Rate5 (highest rate) v Encoder/Decoder complexity ß Encoder ÿ CS1 : 9.27 times more than JM16.1 (4.64 times if implemented on the latest JM) ÿ CS2 : 18.73 times more than JM16.1 (9.37 times if implemented on the latest JM) ß Decoder ÿ CS1 : 7.69 times more than JM16.1 (2.20 times if implemented on the latest JM) ÿ CS2 : 7.11 times more than JM16.1 (2.03 times if implemented on the latest JM) ß Comparison between JM11.0 and the latest JM ÿ JM11.0 encoder is about 2 times slower than the latest JM ÿ JM11.0 decoder is about 3.5 times slower than the latest JM Conclusion v LG’s proposed model is based on AVC/H.264 JM 11.0 v Outperforms alpha anchor and gamma anchor, achieving 25.83% and 37.01% bit rate reduction under CS1 and CS2 respectively v Complexity reduction was not considered in the submitted model. v Coding performance is expected to be further improved if the additional tools are implemented (AWR, PAIF, MVC with B skip/direct, Chroma estimation mode with phase shift)

Score: 677947.95 - https://www.itu.int/wftp3/av-a...10_04_A_Dresden/JCTVC-A110.pdf

  Joint Collaborative Team on Video Coding (JCT-VC) Contribution

In the current TMuC, the Angular Prediction is applied to the PUs of size 8x8 and ADI for the rest of the PUs. (...) The unified design consists of: 1. Prediction directions as defined in ADI with angles of +/- [0,2,5,9,13,17,21,26,32] / 32. 2. (...) Prediction of the most likely direction as in ADI. 3 Objective Performance In order to assess the objective performance of the harmonized intra prediction the following simulation conditions were followed.

Score: 677812.74 - https://www.itu.int/wftp3/av-a..._07_B_Geneva/JCTVC-B100_r2.doc