Organisation internationale de normalisation
Yüklə 2,74 Mb.
|
3Development of AVCThe video subgroup jointly approved the ISO standard related output documents that were produced during the 26th JVT meeting which was held in parallel. Important work items in this context were as follows
The report of the SVC verification tests was finalized, using conditions suitable for a range of possible application scenarios for progressive video, including
For the performance evaluations, SVC was compared against AVC single layer coding by means of subjective testing. Subjective tests were performed following relevant international recommendations using a controlled environment and a high number of test subjects. The results of these tests indicate that these various types of scalability for these applications can be achieved with a bit rate overhead typically equal to or less than 10% when compared to AVC single layer coding using only the highest resolution in the test case. In the HDTV and movie cases, comparable quality was achieved with no apparent need to increase bit rate at all. The bit rate savings obtained by SVC compared to AVC simulcast transmission depend on the particular test case, and were found to be between 17% and 40% of the simulcast bit rate. These bit rate savings relative to simulcast are particularly important for applications in which video must be provided with different spatial resolutions, for which simulcast would previously have been the only available AVC-based standardized solution. All results and more detailed description of the test setup are included in the public test report (N9577). MVC has reached the level of FPDAM as amendment 1 of the new edition of AVC. The specification does not include any new coding tools at the macroblock level or below (see JVT report). A first contribution related to MVC profiling was discussed jointly with JVT and MPEG Requirements SG. The current idea is defining only on “multiview high profile”, no interlaced coding, constraint set flag could be used to perform switching between main and high. Level definitions are preliminary and will need more careful investigation about buffer sizes, restrictions in inter-view prediction etc. In general, it should be avoided to define an entirely new set of levels beyond the existing ones. One solution could be made e.g. by starting from maximum number of macroblocks per second, and derive therefrom useful values e.g. for maximum number of views etc. within a given level. Further study will be necessary taking into account requirements of certain applications (such as stereo, n-view), and investigate for specific levels, whether e.g. the numbers of possible reference pictures are sufficient for the multiview application. Ways to enable parallel processing should also be considered. Documents reviewed:
Documents approved:
Yüklə 2,74 Mb. Dostları ilə paylaş:
|