Thomson recently released its new Tiger software-based compression system, designed specifically for high-definition, MPEG4-AVC encoding. It's designed for PC clusters and is made up of a variety of applications developed by engineers in Burbank, CA and Princeton, NJ. It is targeted at post-production facility "compressionists"- that is, those responsible for managing image quality and bandwidth within a program stream for major movie titles being released on next generation Blu-ray and HD DVD discs. Among the new features are: an advanced GUI optimized for workflows such as the ability to adjust pre-processing scene detection based on groups of pictures; multi-pass variable bit rate allocation, which automatically adds bits to the scenes or frames, over many passes; correction for multiple anomalies, including edges, rain blocking and motion artifacts; and unique technology to accurately replicate film grain after compression. Cooper is general manager of Thomson Technology’s Corporate Research lab in Princeton, and he led the bicoastal team responsible for Tiger’s creation. Princeton engineers designed Tiger's core algorithms, while a Research team in California designed the software's advanced GUI, with input from the company’s co-located Services division.
Q: How do you see the role of the HD compressionist?
A: Within Thomson we refer to a compressionist as someone who is responsible for doing the compression and encoding for different applications, including DVD mastering and replication. Their role is not only to be experts on how to operate an encoder correctly to get the best video quality, but also they should have a golden eye to be able to discern artifacts in an image. Finally, they need to know how to adjust the encoder’s parameters to correct or fix them.
Q: MPEG-4/AVC is particularly challenging to encode. Why?
A: It's got a lot more tools to use within the spec as compared to MPEG-2. It's doing a lot more math calculation, but also it includes advanced features for things like motion estimation. MPEG-2 motion estimation worked in 16×16 blocks, while within H.264 [MPEG-4 AVC] you can look at 16×16, 8×8 and 4×4 blocks, or even non-square 16×8/8×16 or 4×8/8×4 blocks.There's a lot more flexibility and ability to make finer adjustment to the picture when encoding than we had with MPEG-2.
Q: It's interesting that Tiger helps put back film grain, when many in the digital domain have been trying to reduce its effect as much as possible during compression. Why would someone want to add more film grain to an image?
A: What we're trying to achieve with our technology is what we call “transparent” video quality, whereby everything in an image is preserved, including the film grain and the high-frequencies textures, which sometimes can be mistaken for video noise. To build an AVC encoder with those requirements was challenging. We're hitting bit rates in the mid-teens, depending on the title and how much bonus footage is included on a disc. We have run tests down to 10 Mbps, with exceptional quality, but today most titles are targeting 15- 20 Mbps. The Blu-ray and HD DVD specs run up to 40 Mbps. But the average bit rate that content distributors need to include a two-hour movie and all of the extras they are now delivering is 15-20 Mbps. The lower the bit rate, the more content you can get on a DVD disc.
Film grain is often part of the aesthetic of a film, where the director or producer wants to maintain a certain look. That's why people use specific film stocks over others. During the encoding process, standard video compression technologies recognize some of that grain as noise (high frequency) and, as a result, grain tends to get minimized or removed to save space. In contrast, we were very careful in developing our algorithms so that we preserve the grain, if required, and so that compressionists can manage grain, frame by frame. Tiger includes the flexibility to look at specific scenes or frames and makes adjustments based on the compressionist’s preferences.
