Organisation internationale de normalisation

No.	Title	TBP
Available
9568	Disposition of Comments on ISO/IEC 14496-4:2004/PDAM 31	N	08/01/18
9569	Text of ISO/IEC 14496-4:2004/FPDAM 31 Conformance Testing for Scalable Video Coding	N	08/02/29
9572	Disposition of Comments on ISO/IEC 14496-5:2001/PDAM 19	N	08/01/18
9573	Text of ISO/IEC 14496-5:2001/FPDAM 19 Reference Software for Scalable Video Coding	N	08/03/20
9574	Text of ISO/IEC 14496-10:200X/DCOR 1	N	08/04/11
9575	Disposition of Comments on ISO/IEC 14496-10:200X/PDAM 1	N	08/01/18
9576	Text of ISO/IEC 14496-10:200X/FPDAM 1 Multiview Video Coding	N	08/02/15
9577	Report on SVC Verification Tests	Y	08/01/18
9578	Joint Multiview Video Model (JMVM) 7	N	08/02/15
9579	JMVM 7 Software	N	08/02/22
9580	Overview of Multiview Video Coding (MVC)	Y	08/01/18

4MPEG-7 Visual

4.1MPEG-7 Visual related work in Antalya

The MPEG-7 breakout group was active during the whole week. Input documents related to the Visual part in 15938-3 are listed in the table below. All these documents were reviewed and discussed.

m15103	CE Report of VCE-5	Sangyoun Lee
m15104	Text/Logo Mask Image Generation Software for VCE-7	Kota Iwamoto, Ryoma Oami
m15106	Contribution of video test material for MPEG-7 video signature CE	Weon Genu Oh, Daeil Yoon, Jie Jia, Hae Kwang Kim
m15114	Proposal on Frame-Reduction video clip format	Ju-Kyong Jin, Weon-Geun Oh, Dong-Jin Seo, Sang-il Na, Jae-Hyun Huh, Dong-Seok Jeong
m15122	Errors in MPEG-7 reference software	James Annesley
m15130	Experiment Results of Image Signature for Complex Conditions	Weon-Geun Oh, Ayoung Cho, Won-Keun Yang, Ik-Hwan Cho, Ju-Kyong Jin, Jun-Woo Lee, Dong-Seok Jeong
m15131	The Extra Experiment Result to Verify Performance Measure Method of MPEG-7 VCE-6	Weon-Geun Oh, Won-Keun Yang, Ayoung Cho, Dong-Seok Jeong
m15137	Cross verification result for ETRI VCE-6 proposal	Min-Jeong Lee, Heung-Kyu Lee
m15139	Video Signature based on Inter-frame Correlation Coefficients	[e-mail] Zheng Huang, [e-mail] Tiejun Huang, yhtian@ @pku.edu.cn Yonghong Tian
m15140	Visual Signature based on Waston Perceptual Model	[e-mail], [e-mail], [e-mail]
m15169	Correction to Image Signature XM Software	Paul Brasnett, Miroslaw Bober
m15170	Performance Evaluation of Image Signature on Extended Database	Paul Brasnett, Miroslaw Bober
m15172	Extending the Trace Transform Image Signature to Complex Conditions	Paul Brasnett, Miroslaw Bober
m15181	Cross verification result of Image Signature (VCE-6)	Karol Wnukowicz
m15217	Updated Results on Extended Trace Transform Image Signature	Paul Brasnett, Miroslaw Bober

On major work item has been the further review of image signature descriptors as investigated in VCE-6. A Dataset of approximately 130.000 images was used. Independence was tested on 8.45 billion image pairs. Robustness was tested on 250.000 images (24+1 different modifications). The following findings were made, comparing 15130 vs. 15217:

• 
  Average performance above 90% at 10ppm for both methods
  
• 
  15130 (“concentric circles” method) performs significantly better for cropping
  
• 
  15217 (extension of method in XM/WD, “trace transform” method) performs slightly/sometimes significantly better for most other cases
  
• 
  Cropping result is very specific, because 15130 method would fail in case of non-center crops, but 15217 has also very poor result for case of heavy cropping
  

In general, it can be concluded that both methods require more development for the cases of the more difficult modifications. In particular, further investigations appear necessary for cases of more localized signatures (which were extracted similarly in both methods, using areas around feature points). The global descriptors appear useful only for simple conditions, but could be also used as first step in quick database search, sorting out the clearly dissimilar images.
The following decision was made:

• 
  The technology of current WD (global signature) is promoted for PDAM
  
• 
  For complex operation, most probably good localized descriptors would be required, which must not necessarily be derivates from the current global descriptor. Further investigations will be made in the upcoming CE, also taking into account combinations of complex conditions, such as combination of translation and cropping (no-center case), and also with scaling. necessary
  
• 
  The method of 15217 (global WD/XM descriptor with some more localized feature extraction) will not be put into a new version of the XM, because it may turn out in CE that other localized descriptors perform better.
  

For Video Signatures, it is estimated that 208 hours of video content are required, of which ~100 hours have already been collected or committed to be submitted in near future. Varied content is required such as: sports, news, film, soap, variety of others. The set should consist of approx. 4.000 longer clips and 24.000 shorter clips; 50 million comparisons must be performed for the envisaged range of quality (successful hits vs. false alarm rate).

Furthermore, work on software to create video modifications automatically has been done in VCE-7. Commitment for providing the remaining data material is expected for the time between Antalya and Archamps meetings. It is therefore decided to delay the CfP by one meeting to finalize the testing database work.
Technically, two new proposals (ideas) were received and reviewed, related to video signatures. Their testing awaits completion of the full database.
The following timeline is planned for the ongoing work on video signatures:

• 
  Final CfP: 2008/04 with responses 2008/07
  
• 
  PDAM: 2008/07
  
• 
  FPDAM: 2009/01
  
• 
  FDAM: 2009/07
  

Three Core Experiments will continue:

• 
  Face Recognition in IR images
  
• 
  Image Signature for Complex conditions
  
• 
  Video Signature (collection of test material and software tools for CfP preparation)
  

A bug report concerning visual XM was reviewed. The bug reported and solutions proposed will be verified by the XM maintenance team at Warsaw University.
The development of the reference software for ISO/IEC 23000-3 was continued. The BIM incompatibility issue was solved, new SDL’s were provided to generate bitstreams. The Software is being rewritten to reflect new SDL’s, expected completion is within 3 weeks after the Antalya meeting.