A closing plenary of the Video subgroup was held Thursday 16:00–18:30 and continued Friday 8:00. Output docs were approved, and AHGs were established. Discussions are recorded in this section.
All issues regarding breakout activities are basically reflected in the respective subsections about BoG activities above.
CDVS:
One change in the keypoint detection – before computing SIFT
Definition of conformance: Two options: Accuracy of keypoint detection for a given set of images, criterion is similarity with the “golden response” of the reference. This could be different for the global and local descriptor.
Software: Concern was expressed by developers about changing to the BSD copyright statement. The main concern is 1) the change of the ownership 2) that perhaps the original contributors would not be allowed to re-pupose the code for non-standard purposes. Likely, the second aspect does not apply. Further clarification necessary, needs to be resolved until the next meeting. One question is what “contributor” means in the header, this is currently anonymous. If it cannot be resolved, a second implementation could be necessary.
Awareness event is planned for Warsaw.
IVC:
First results were presented, but some concern was expressed regarding large confidence intervals and some unreasonable results (saturation of quality over several rate points).
The results were corrected later by the test chair. It was requested to make an initial analysis (by Ronggang Wang) until Friday morning, with a similar scoring scheme as was applied in San Jose.
Other topics in closing, as appropriate, are described under the respective sections from the BoGs.
On Friday morning, it as requested to start a new work item on IVC in the video plenary. However, by that time, no preliminary analysis of the results of the evaluation had been available yet. An initial inspection shows that the quality of ITM10 would be roughly comparable to WVC and VCB on various cases. The test report N14989 requires more analysis together with the test chair, also taking into account the rate points for some of the test cases and will have an editing period of three weeks.
In terms of a request for subdivision, it would be necessary to justify that IVC has properties that the other two codecs cannot fulfill. This could for example be demonstrating from the test results that the quality is better than the other two codecs, that it has different application areas etc.
From the perspective of the video subgroup, it is impossible to come to a decision at this late stage of the current meeting. Further considerations should be made by the next meeting after availability of the report.
The Video plenary was closed Friday 24 Oct., 09:36.
