International organisation for standardisation organisation internationale de normalisation



Yüklə 5,72 Mb.
səhifə44/84
tarix25.12.2017
ölçüsü5,72 Mb.
#35931
1   ...   40   41   42   43   44   45   46   47   ...   84

8TE6: Intra prediction


8.1.1.1.1.1.1.1.1JCTVC-C172 TE6: Summary report of intra prediction improvement tool experiments [A. Tabatabai (Sony)]

This contribution was a summary of Tool Experiment 6, Intra Prediction Improvements. From a total of 6 proposed toolsets, 4 had been evaluated by at least one organization. For cross checking, the recommended test conditions of intra-only were used for both high compression efficiency and low complexity as defined in the document JCTVC-B306_r3.

The tools tested in this tool experiment were as shown in the table below:


Tool

Technology

Proponent

Cross-checker(s)

TE6.a.1

Weighted Bidirectional Prediction (WBP) and Adaptive Sub-block Coding Order (ASCO)

Toshiba

JCTVC-B042

JCTVC-C079


Sharp JCTVC-C179

Renesas JCTVC-C187



TE6.a.2

Short distance intra prediction

Huawei & Hisilicon

JCTVC-B040

JCTVC-C101


LGE JCTVC-C028

DOCOMO JCTVC-C045

Ericsson JCTVC-C157

Microsoft JCTVC-C271



TE6.a.3

Overlapped block intra prediction

MediaTek

JCTVC-A109

JCTVC-C193


Microsoft JCTVC-C272

TE6.a.4

Line-based intra prediction

Microsoft

JCTVC-A118

JCTVC-C270


Huawei JCTVC-C139

DOCOMO JCTVC-C045



TE6.b.1

Differential coding of intra modes (DCIM)

Sony

JCTVC-B109

JCTVC-C169


Sharp JCTVC-C175

Panasonic JCTVC-C217





8.2TE6.a.1 Weighted Bidirectional Prediction (WBP) and Adaptive Sub-block Coding Order (ASCO)


8.2.1.1.1.1.1.1.1JCTVC-C079 TE6 subset a: Bidirectional intra prediction [T. Shiodera, A. Tanizawa, T. Chujoh, T. Yamakage (Toshiba)]

This contribution presents the detailed experimental results of Bidirectional Intra Prediction (BIP) for Tool Experiment 6 on intra prediction improvement. BIP was included in the CfP submission of JCTVC-A117 and was proposed in the contribution of JCTVC-B042 at JCT-VC Geneva meeting.

In this document, experimental results using TMuC software under the common test conditions defined by JCTVC-B300 were reported. For the I slice only coding structure, the average BD-rate gain was 2.1% on low complexity conditions and 2.3% on high efficiency conditions without increase of the encoding time and the decoding time.

The reason that the encoding time was not increased was noted to be fast mode decision and turning off some tools that were enabled in the reference configuration – adaptive intra smoothing (AIS) and edge-based prediction. This feature would ordinarily be expected to increase encoding complexity.

It was remarked that JCTVC-C207 encoding optimization improves efficiency by 0.8% with only a fast mode decision technique without adding new tools.

Further study was encouraged.

8.2.1.1.1.1.1.1.2JCTVC-C179 TE6: Cross-check of bi-directional intra-predition results from Toshiba [A. Segall (Sharp)] (missing prior, available first day)

This contribution cross-checks the bi-directional intra-prediction results provided by Toshiba. This cross-check was performed within the context of TE6 for "Intra high-performance" and "Intra low-complexity" configuration settings. The experimental results reportedly closely matched the rate-distortion results provided by Toshiba. The maximum Y BD Rate differs from the provided results by at most 0.1%.

The Toshiba software source code was used, and the software was reportedly reviewed in detail.

8.2.1.1.1.1.1.1.3JCTVC-C187 TE6.a: Cross-check of bidirectional intra prediction (JCTVC-B042) [K. Iwata (Renesas)]

This document described results of cross-checking the software containing the tool of bidirectional intra prediction (BIP) proposed by Toshiba Corporation. The verification task was reported to have been done successfully and the results matched with proponent’s data. The Toshiba software source code was used (without studying the software algorithm).

8.3TE6.a.2 Short distance intra prediction


8.3.1.1.1.1.1.1.1JCTVC-C101 TE6.a: HiSilicon report on short distance intra prediction [C. Lai, Y. Lin (HiSilicon)]

This document reported the Tool Experiment TE6.a activities related to the evaluation of short distance intra prediction (SDP). Based on the test results achieved so far, line based SDP was reported to provide about 0.9% of bit rate reduction over TMuC anchors for the Intra LC configuration with comparable encoding and decoding time.

The proposed tool includes two kinds of 1x16, 16x1, 2x8, 8x2 and "resample based" prediction.

The "resample based" prediction was not tested, and the HE configuration context was not tested due to software development difficulties.

It was remarked that the LC configuration context may not be the best in which to evaluate such a technique, since it seems to have substantial complexity impact.

Further investigation was encouraged.

8.3.1.1.1.1.1.1.2JCTVC-C028 TE6: Verification results of Huawei and Hisilicon proposals (short distance intra prediction using the correlation between lines and pixels) [J. Kim , B. Jeon (LG Electronics)]

This document reported verification results of JCTVC-C101 on short distance intra prediction. The proponents provided their source coded which was compared under the condition of intra low complexity condition, and the results reportedly matched those of JCTVC-C101.

8.3.1.1.1.1.1.1.3JCTVC-C045 TE6.a: Cross verification for line based intra prediction [T. K. Tan, Junya Takiue (NTT DoCoMo)]

This contribution reported simulation results and cross verification of subset Tool Experiment 6.a as described in JCTVC-B306. This contribution documents the findings based on the study and simulation conducted using the software provided by Huawei and Microsoft.

This contribution was a cross-check verification for both JCTVC-C101 and JCTVC-C270.

Software was received from the proponents, and was somewhat studied to identify the algorithmic differences. In the case of JCTVC-C101 (JCTVC-B040) verification, the software changes were more difficult to identify and study, due to not being controlled by a macro switch.

The results reportedly closely matched those reported for JCTVC-C101 and JCTVC-C270.

The contributor remarked that deeper analysis of the changes made in these proposals in regard to the prediction process and to the difference coding process was desirable.

The contributor also remarked that it would be beneficial to study harmonization of JCTVC-C101 and JCTVC-C270 as well as refinement of each scheme itself.

8.3.1.1.1.1.1.1.4JCTVC-C157 TE6.a: Cross verification of line-based intra prediction [J. Samuelsson, R. Sjöberg, K. Andersson (Ericsson)]

This contribution contained cross-checking results of Line-based Intra Prediction (LIP) presented by Huawei in JCTVC-B040 and later implemented in TMuC v7.0. Simulations run for the Intra Only Low Delay test case reportedly showed an average BD-rate reduction of 0.9%, matching what was reported in JTCVC-C101.

The software was studied to determine that it matched the described algorithm.

8.3.1.1.1.1.1.1.5JCTVC-C271 TE6: Cross-verification of HiSilicon's short distance intra prediction by Microsoft [X. Peng, J. Xu (Microsoft)]

This document summarized the Tool Experiment TE6 activities related to the cross verification of short distance intra prediction by HiSilicon.

The source code provided by HiSilicon was reported to match with their document description, except that the resample-based intra prediction was disabled. Only intra prediction by exploring the correlations between lines was enabled. In prediction, four modes were allowed on 16x16 and 8x8 blocks, which are 1x16, 16x1, 2x8 and 8x2. Each line had its own prediction mode among nine, which needs to be coded in the bit stream. After prediction, a DCT is applied on the line before it is quantized.

The results reportedly essentially matched those reported in JTCVC-C101.



Yüklə 5,72 Mb.

Dostları ilə paylaş:
1   ...   40   41   42   43   44   45   46   47   ...   84




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

gir | qeydiyyatdan keç
    Ana səhifə


yükləyin