The following input documents had initially been reviewed in the AHG 14-01-12 Meeting and further discussions were performed in the BoG during the meeting week.
14.1.97.1.1.1.1.1.52m32144 Performance evaluation of DC intra prediction mode for Internet Video Coding [Dong-Hyun Kim, Haechul Choi, Jae-Gon Kim]
Summary
In this document, the performance of ITM 7.0 intra prediction in which five-tap smoothing filter was used for DC intra prediction in comparison with three-tap smoothing filter was presented.
The five-tap smoothing filter provided 0.1% bit saving in average over the three-tap smoothing filter in CS1 encoding configuration without complexity increase.
Comments
This contribution was an informative document.
Q: Is there any increase of computational time in en/decoder?
A: The proponent said that there was no increase of computational requirements because of the portion of intra prediction.
14.1.97.1.1.1.1.1.53m32455 Cross-check of m32144 [Kiho Choi, Minsoo Park, Sang-hyo Park, Euee S. Jang]
This contribution was a cross-check report of m32144.
The software provided by the proponent was reviewed and was consistent with the descriptions in m32144.
14.1.97.1.1.1.1.1.54m32145 Performance evaluation of non-reference P frame coding in ITM 7.0 [Dong-Hyun Kim, Byung Tae Oh, Jae-Gon Kim]
In this document, the performance of non-reference P frame in ITM 7.0 was presented as a part of comparative investigation of Internet Video Coding technologies.
The non-reference P frame coding provided up to 5.8% bit rate saving on average over the usual low-delay encoding (CS2).
14.1.97.1.1.1.1.1.55m32456 Cross-check of m32145 [Kiho Choi, Minsoo Park, Sang-hyo Park, Euee S. Jang]
Summary
This contribution was a cross-check report of m32145.
The software provided by the proponent was reviewed and was consistent with the descriptions in m32145.
A resolution-dependent interpolation filter for different resolution videos was proposed.
In this contribution, a 4-tap filter is proposed for the UHD (2560×1600 and above) videos, a 6 tap filter was proposed for the videos with resolution ranging from 720P (1280×720) to 1080P (1920×1080), and a 10 tap filter is proposed for the videos with the resolution below 720P, respectively.
Simulation results showed that the proposed scheme provided 1.3% bit rate saving in CS2 and 0.9% bit rate saving in CS1.
The analysis of complexity showed that the increased computational complexity was relatively not so huge compared the use of 6-tap filter.
The prior art information based on US patent database was provided in the contribution.
Comments
Q: Do you have the result of the computational time using the proposed method?
A: Slightly increased. Nevertheless, the proponent explained that the increased time was not huge in principle.
Q: Did IVC have a plan to support an UHD application?
A: According to the IVC requirement document, it was not planned.
Q: How do you generate the coefficients for the filter?
A: The generated coefficients was based on that in a text book. The information of filter generation was not described in the contribution. The information will be added on the contribution during this MPEG meeting.
Q: How to signal to decoder?
A: The proponent said that there was no an additional signal but the method utilized the syntax used for a sequence resolution.
Q: Did you find a similar syntax in your investigation for prior arts?
A: No.
Q: Did you study JCT-3V contributions? There would be a similar syntax.
A: The proponent studied the JCT-3V contributions but the there was no contribution exactly same with the proposed method.
Q: Why not sending the syntax directly?
A: The proponent explained that there would be an IPR issue.
Q: Do you check the prior art in terms of adaptive interpolation?
A: Needed to check.
The proponent used an existing syntax in ITM 7.0, which was for the information of sequence resolution, in order to signal which interpolation filter was used.
Experts in the group concerned that the use of the existing syntax was one of ways to signal to decoder.
The proponent believed a signalling using an existing syntax was safe about the IPR issue because the existing syntax was not for the purpose of the proposed method.
Agreement
Technically, there was no problem to adopt the method. However, the issue with regard to the use of existing syntax, which was used for another purpose, has not been studied.
Further study on the IPR issue during this meeting.
Report this issue on the video group meeting.
14.1.97.1.1.1.1.1.57m32175 Cross-check of m32173 on interpolation filter [Weiran Li, Yun He]
This contribution was the cross-check report of m32173.
The software provided by the proponent was reviewed and was consistent with the descriptions in m32173.
In this contribution, 16x16 block size was proposed to improve coding efficiency of IVC.
Simulation results showed that the proposed method provides 2.2% bit rate saving in CS2 and 2.6% bit rate saving in CS1.
The prior art information based on published papers was provided in the contribution.
This contribution was based on a paper issued in 1989.
Comments
The proponent said that the proposed method used the same 16x16 kernel regardless of intra or inter block but the signal to indicate the use of 16x16 kernel was used only in intra block.
One expert said AVC/H.264 uses 16x16 kernel for a transform process.
One expert said there would be possibility on an IPR issue when variable block sizes were used with intra prediction. The year when the use of intra prediction was introduced has not been passed over 20 years. Thus, the expert concerned the use of the proposed method with other techniques.
One expert concerns about the increase of en/decoding time
Agreement
The group believed that there was no problem to adopt the method in the aspects of a technic merit as well as an IPR issue related to the use of transform sizes. Nevertheless, the group members were concerned about the IPR issue when the technology combined with the intra prediction.
Further study on the IPR issue during the MPEG meeting.
Report the issue on the video group meeting.
14.1.97.1.1.1.1.1.59m32176 Cross-check of m32174 on ABT [Weiran Li, Yun He]
Summary
This contribution was a cross-check report of m32176.
The software provided by the proponent was reviewed and is consistent with the descriptions in m32176.
14.1.97.1.1.1.1.1.60m32445 Deblocking Method for Internet Video Coding [Minsoo Park, Kiho Choi, Jin Yeon Choi, Sang-hyo Park, Euee S. Jang]
Summary
In this contribution, a deblocking method, which was applied to in-loop process, was proposed.
The experimental results showed that the proposed method improved the visual quality of ITM 7.0 significantly without any coding loss.
The proposed method was based on a paper issued on April 1994.
Comments
Q: Why is the de-blocking filter just applied to intra blocks with DC coefficient only?
A: Applying to inter coding, the improvement on inter block is little.
Q: How about the complexity increase of the filter?
A: since it is only applied to intra blocks with DC coefficient only, this kind of blocks are few, the complexity increase is very limited.
Agreement
Further discussion during this meeting
14.1.97.1.1.1.1.1.61m32454 Cross-check of Deblocking Method for Internet Video Coding (m32445) [Dong-Hyun Kim, Jae-Gon Kim]
Summary
This contribution is a cross-check report of m32445.
The software provided by the proponent was reviewed and is consistent with the descriptions in m32445.