The business of testing very high compression bit rates at higher frame rates is not so simple. Both content and testing apparatus require special consideration, as needless to say, this has never been done before.
It is the visual quality of JPEG2000-compressed images at different compression rates that must be studied, which renders the litany of test charts and test sequences developed for MPEG-based broadcast content to be of no use. This is due to the nature of JPEG2000, which is based on the discrete wavelet transform, while MPEG is based on very different type of algorithm called the discrete cosine transform. In practical terms, this means that JPEG2000 images degrade differently from their MPEG counterparts. As the bit rate falls, sharp edges will go soft from ringing, and detail throughout the image begins to blur. JPEG2000, as applied in digital cinema, only operates frame-by-frame, and does not take into account the history of the image, as is commonly done in MPEG. Motion artifacts, therefore, are not systemic, but instead a problem introduced through ringing and blur in each individual frame. The purpose of compression testing is to learn the bit rate above which diminishing returns occur. To do this, test content must be specially constructed to bring out the worst.
With content in hand, it must be compressed at various bit rates, and then projected in a screening room. Your author, as co-chair of the SMPTE Study Group for Higher Frame Rates, has taken the approach of extending existing standards. It’s no great exercise for available digital cinema packaging tools to compress the content and package it in a DCP. Where the muscle is required is in the playout of the DCP. No off-the-shelf equipment today can support the necessary 1Gb/s to 1.5G/s decompression rates that are needed to properly conduct the compression tests. This is not to say that media blocks in the cinema must decode 1Gb/s to 1.5Gb/s bit rates. The expectation is that cinema will require less. But unless a higher bit rate is observed, no one will be able to say for certain where diminishing returns occur.
On the content side, the Study Group is fortunate to have expert cinematographers step in to assist. Doug Trumbull has volunteered 3-D content now being shot at 120 fps per eye, which can be merged to produce a 60 fps per eye test piece. If all of the moving parts fall in place, Doug’s content may be shown at NAB, a week prior to CinemaCon. David Stump, also co-chair of the SMPTE Study Group, has volunteered to shoot specialized test sequences, using equipment and facilities volunteered by Cameron-Pace. Cinematographer Kommer Kleijn, the third co-chair of the SMPTE Study Group, is participating in the design of the specially produced test sequences.
The playback system requires equally detailed attention. Jean-Francois Nivart, one of the original collaborators in the design of the DCP, has founded a new company, Image Matters, intended to solve such implementation problems. Using the PCIe-based digital cinema media block development system offered by Intopix, and PCIe-based high speed flash storage from Virident, a system is under construction to perform the high speed decompression needed. Notably, there are no standards for communicating 120 fps over multiple 3G-SDI links to the projector. More collaboration is needed, and media block manufacturer Mikrom is working with Image Matters in the coding of image data between processor and projector. If all goes right, the only thing special about the projector will be the high frame rate upgrade now available to cinemas.
Given the degree of volunteer work that these efforts require, the schedule for this work will take a good part of this year to complete. Unlike the original digital cinema compression tests held only in Los Angeles in 2000, the higher frame rate tests will be conducted at multiple sites around the world. Currently, test sites in Belgium, New Zealand, and Burbank are planned. The result of these tests will have a significant impact on the future design of digital cinema equipment, making these tests worth doing well.