| Posted: 22 Dec 2011 01:33 PM PST JPEG 2000 has caught the attention of the professional media world for good reason. First, it closely matches some workflows, where the production process operates on each frame of a video stream as a discrete unit. This is different from MPEG-2 and MPEG-4 AVC Long-GOP flavors, where, during the reconstruction process, algorithms reference frames before and after the frame being reconstructed. The ability to compress each frame as a free-standing unit has made it popular in the digital intermediate space in Hollywood. JPEG 2000 is also of interest to those who want lossless compression. It can provide a bit-perfect reconstruction of the original compressed image, although at a cost in terms of bandwidth. Also, the wavelet compression used in JPEG 2000 provides some unique opportunities that are not available in other compression methods. The wavelet transform separates the image into four sub-bands. The first sub-band is a lowpass horizontal and lowpass vertical (LL). Images that have gone through this separation are basically lower-resolution images of the original. The other sub-bands are as follows: lowpass horizontal and highpass vertical (LH); highpass horizontal and lowpass vertical (HL); and highpass horizontal and highpass vertical (HH). Using wavelet transforms and some clever thinking, implementers can do some interesting things. For example, they can send only the LL image, if they know that they are feeding a low-resolution display. Or they can send the LL sub-band in a highly-protected stream, in order to ensure the original image arrives intact. That said, they can then send the higher-resolution sub-bands unprotected since a momentary loss of these sub-bands is not likely to be noticed. Given JPEG 2000's popularity, it is not surprising there have been some developments that make it particularly interesting for professional applications. First, the ITU has created an amendment (1) that outlines specific configurations for broadcast contribution applications. These configurations are intended to establish interoperability points for those implementing JPEG 2000 in professional applications. This is important because, until the amendment was released, there were so many variables in the compression tool set that interoperability was unlikely. The second important development, Amendment 5 to the MPEG 2 standard (2), provides a mapping of the JPEG 2000 Program Elementary Stream (PES) onto the MPEG-2 Transport Stream (TS). Finally, some time ago, the Pro-MPEG Forum started to develop a standardized way to transport MPEG-2 TS over IP networks. The Video Services Forum picked up on this work and continued to develop it, finally submitting a draft for standardization within the SMPTE. This standard, SMPTE 2022-2 (3), describes a method for mapping MPEG-2 Transport Streams onto IP networks using RTP and UDP. The document was approved in 2007 and is the most common standard deployed today for professional video transport applications. So these three developments — development of broadcast profiles; a mapping of JPEG 2000 Program Elementary Streams to MPEG 2 Transport Streams; and wide availability of MPEG-2 TS over IP transport equipment — mean now it is possible to transport JPEG 2000 over IP networks.  broadcast contribution profiles, JPEG 2000 PES to MPEG-2 mapping, and SMPTE 2022-2 for MPEG-2 transport over IP networks. Starting with a video source, the image is compressed using a compression engine. This engine is configured to one of the Broadcast Contribution profiles in ITU-T Amendment 3. The compression engine produces a JPEG 2000 PES. This stream is then fed to an MPEG-2 encapsulator. The encapsulator uses the mapping rules established in the MPEG-2 specification, Amendment 5, to map the PES onto an MPEG-2 TS. This MPEG-2 TS is fully compliant with MPEG-2 specifications because, from the outside, it looks just like a normal MPEG-2 Transport Stream. As such, the output of the MPEG-2 encapsulator can be fed into a SMPTE 2022-2 compliant video transport device. This device encapsulates the MPEG-2 TS in standard RTP and UDP packets, and then those packets are wrapped in IP packets. These IP packets can now be fed into an IP network. You might wonder why we take a relatively new compression algorithm such as JPEG 2000 and encapsulate it in MPEG-2. There are several reasons for this. First and foremost, there are already a number of specifications for how to encapsulate a number of different audio formats into MPEG-2 Transport Streams. Remember: JPEG 2000 says nothing about audio. Using MPEG-2 TS allows us to transport and present the audio alongside the JPEG 2000 compressed video using well-known audio standards. Also, this approach allows us to leverage the existing SMPTE 2022-2 MPEG-2 TS over IP standard. Finally, there are no technical issues in MPEG-2 TS that need to be fixed in this application space, so re-use of Transport Streams rather than inventing something entirely new seems like a good solution. So, the good news is that the time is ripe for development of an interoperable, open solution for the transport of JPEG 2000 video and audio over IP networks. The standards exist, and there is a clear path forward. But, there are a few issues that need addressed. JPEG 2000 has been around for quite some time. As such, some proprietary JPEG 2000 over IP transport solutions have already been created. Of course, these were developed in response to customer demand, so existing implementations may need to be changed. Another issue is that while the broadcast contribution profiles in Amendment 3 go a long way toward interoperability in the JPEG 2000 PES space, recent analysis suggests, without further definition, implementations based upon these profiles will not be interoperable. Finally, until the industry actually tries to connect devices from different manufacturers together, interoperability cannot be assured. Fortunately, the industry is becoming aware of these issues, and steps are being taken to begin work in earnest on interoperable, open transport of professional JPEG 2000 images over IP networks. I would expect to see some developments around this in the first half of the coming year. Footnotes:
By Brad Gilmer, Broadcast Engineering   |
Friday, December 23, 2011
JPEG 2000
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment