Multimedia
Blogs
Forum
Affordable HD
Whitepapers
Newsletters
Events
Advertisers
DCP Directory
MillimeterA DI Primer
Mar 1, 2004 12:00 PM, By Matt Cowan Loren Nielsen
Digital Intermediate (DI) is a general name that covers a wide range of processes related to the digital postproduction of motion picture films. While presented these days as a new workflow, most individual parts of the process have been around for a long time — scanning, digital effects, compositing, digital editing, and film recording. What's new is the degree of integration of these processes and the ability to evaluate the look of the digitally manipulated images without requiring printing to film. The output of the digital intermediate process, therefore, becomes the master of the movie in digital form. From that master, a film negative is created for the purpose of striking release prints, and/or a compressed or uncompressed digital file is created for digital distribution.
Many vendors will emphasize digital intermediate hardware and software in various iterations at NAB 2004. Along with providing a general overview of the components that, together, make up the DI process, this article offers a basic structure to help those readers shopping for DI tools evaluate different systems and approaches.
Putting a project through a digital intermediate offers both creative and workflow benefits. From a creative point of view, putting an entire feature film into the digital environment provides flexibility in color, dynamics, framing, and image enhancements not easily available via lab processes. The DI workflow also allows CGI shots to be seamlessly integrated into the movie, opticals to be done digitally — with faster turnaround and higher quality — and it offers the potential to make each film print from the same “original negative,” the negative created directly from the digital data. The digital master created by DI processes can also easily be converted to other formats for digital cinema release and home video/DVD applications.
The general digital intermediate process could apply to images that are captured on film, captured in digital cameras, or created entirely digitally. For the purposes of this article, we will focus on film origination, as this represents the major application of the process today.
Components
Let's examine the central aspects of a DI process:
Digital Representation: Central to the DI process is the digital representation of the picture. This representation is the data that forms the basic (input) images, and is manipulated and adjusted to achieve the final look of the film.
The most common data format currently in use is 10-bit log printing density. (This format is generically known as DPX — SMPTE 268 — but perhaps is better known by the Kodak-developed variant, Cineon.) Other formats, such as Open EXR — a 16-bit or 32-bit floating-point representation — or 16-bit SGI or TIFF, are sometimes also used. When the data is being manipulated, for color correction or special effects, it is usually converted to a higher bit depth for performing calculations to avoid quantization and round-off errors.
Resolution for the process is an important aspect of data representation. A 2K file represents approximately 12MB per frame, while a 4K file is four times larger — about 48MB. Hard-disk costs and CPU speeds for processing data are improving rapidly, but practical bandwidth solutions for moving data often create the bottleneck in postproduction, and time becomes a major driver for postproduction costs. Usually, economics wins out, and most digital intermediates are processed at 2K resolution.
Scanning: There are two classes of realtime scanners available — the telecine type and flatbed scanners. Flatbeds generally provide a higher quality, at the expense of slower scans (seconds per frame) and a resulting higher scanning cost per frame.
Scanner performance has several important dimensions. Among them is “dynamic range,” meaning the scanner needs to capture the entire dynamic range of information on the negative. Even though this may contain information that will not be shown in the final image, it gives headroom and footroom to change the dynamics of a shot as required to fit creative requirements.
Stability is also crucial for scanner performance because there cannot be any instability in the scanned images from frame to frame. Pin-registered scanners, designed to be very stable, tend to be slower. Edge-registered scanners may exhibit slight frame-to-frame movement in the scanned image.
Scanning resolution is the other key issue when it comes to scanner performance. Numerous tests and demonstrations have been done to show the ultimate resolution of film. The question of scanner resolution addresses similar questions to the resolution in data representation. From a sampling point of view, it is generally true that a greater number of pixels provides a higher-quality image. Most scanning is currently carried out at 2K resolution, due to additional time and data storage required to handle the picture resolution. 4K scans are sometimes performed, with the scanner immediately down-rezzing the data to 2K before it leaves the scanner. (See the September 2003 issue of Millimeter or millimeter.com for “Digital Cinema's Special ‘K,’”an article examining 2K and 4K resolution processing.) The tradeoffs are scanning time per frame (and cost), data transfer times, and processing costs.
Telecine scans provide a more cost-effective solution because of the speed of the scanning process — 18fps as opposed to the several seconds per frame that a flatbed scanner requires. Many facilities have adapted their telecines to the DI workflow by using them in data mode in which the output is 2K by 1556 pixels, a higher resolution than the normal 1920x1080 used for television, and adequate if not ideal for DI.
Storage, Networking and Data management: There is no DI without storage. However, in addition to the sizeable storage support that any DI system requires, the infrastructure for moving data is equally important. Although you can buy an engineered solution, generally you will need to purchase and integrate your storage and data networking with your chosen scanning hardware. Data management is often overlooked — it's clearly not a glamorous part of DI, but it might be the most important. The DI process lives or dies by the ability to manage the digital assets, control multiple versions, and accurately assemble the final film. Any capable DI process must have a strong asset-management system. Unfortunately, as crucial as this aspect of the process is, it is still also in development around issues of interoperability. Some manufacturers, such as Grass Valley, provide a digital asset-management component with their engineered solutions. Some facilities have developed proprietary asset-management systems as part of their DI workflow. Future progress in this area will be essential to the widespread adoption and cost-effectiveness of DI.
Image Processing: This is the actual manipulation of the various images, which is done in many cases via classic color correctors, compositing systems, or a combination of the two. In addition to performing general color grading, scratch removal, management of image dynamics, and certain special effects, image processing in the DI process also serves to incorporate and unify a variety of CGI shots that may have come from one or multiple facilities. These are sequenced into the movie during the DI, and additional processing is performed to correct color, process grain, add sharpening, and frame the images. The results of image processing bring the desired look to the movie and provide the ability to achieve shot-to-shot consistency.
Realtime Viewing Environment: This is where creative intent and the technology intersect. While the projected film print represents the critical point of approval, the actual look of the film varies slightly depending on the characteristics of the film projector. A stable digital viewing environment — ideally based on a large-format projector but more often combining calibrated monitors and projectors — therefore, provides a consistent review environment, and if well matched to the look of the output film, it provides a point at which filmmakers can approve the look of an image without multiple film-out iterations.
Output to Media: For current theatrical releases, the finished data is usually recorded to film, leading to the creation of a negative for replication of release prints. Because the level of consistency available in DI, the digital master is more consistent than a conventionally timed interpositive, ultimately providing higher quality. This makes the creation of digital cinema and home video/DVD masters more straightforward, and it can usually be done with a simple color-correction trim pass. Work is currently underway to automate this process as much as possible, reducing the time and cost of making these masters.
Key Issues
To get the best advantage from the DI process, several key issues need to be properly addressed. These include:
Reference Image: Currently, film output represents the final creative reference. This is because film is the dominant largescreen output format, at least for the foreseeable future. Film, however, is a difficult medium for reference — color and tonality vary depending on specific lab processes and the color characteristics of the film projector displaying the film. One of the big advantages of performing a DI is that it is viewable before it is recorded to film, and filmmakers can adjust its properties with a clear understanding of how those adjustments will look upon a film output. To achieve this, it is necessary to be able to accurately emulate the look of film on a digital display.
Traditionally, digital representation has been viewed on a CRT display. CRT monitors produced adequate results, but they lacked the color gamut (range) and the image size and presence of a theatrical display. More recently, DLP Cinema and D-ILA displays have found their way into digital color-correction rooms across the world for final image viewing and approval. These projector technologies are very stable and not subject to daily lab differences or film projector color differences. In setting up a DI facility, though, one critical point is the need to calibrate the process to ensure that the final film output matches the look of the digital display.
Color Calibration: Calibration requires closing the loop between the look of the digital image and the film image. To achieve this, calibration lookup tables (LUTs) must be created to translate between the data representation on film and the digital display to make a visual match. These LUTs are usually devised by digitally creating a large number of color patches and measuring the resulting color on film and on the digital screen. This, in turn, allows LUTs that match the digital and film processes, providing a unified look. Sophisticated color matching is usually done with 3D LUTs.
In postproduction, the LUT may be deployed in the color-correction hardware/software or in the projector and film recorder devices. Post facilities work hard on color management, and they often promote their ability to achieve accuracy and quality. Their track record in this regard is often a key issue in luring filmmakers to their facilities for DI work.
System Implementation and Integration: Many manufacturers offer hardware and software components for the DI process, but no company offers a turnkey solution. To date, in actuality, most post facilities design their own custom systems to suit their specific workflow requirements and integrate with existing hardware. Each of them uses different combinations of well-known industry hardware and proprietary tools and techniques.
In any case, however, the DI offers significant advantages to the creative process in filmmaking. The number of films using this process and the number of facilities offering the service are both increasing rapidly. As process issues are ironed out, many more — and even a majority of — feature films will undoubtedly be posted digitally using one type of DI process or another.
To learn more about such DI-related topics as calibration and color dynamics, you can attend Matt Cowan's SMPTE seminar at NAB, Saturday morning, April 17. For more on Cowan's seminar and other SMPTE day events, go to www.nabshow.com/dcs.asp.
Matt Cowan is a principal at Entertainment Technology Consultants (ETC), an organization specializing in the science and applications of digital cinema technology. He has more than 20 years experience in the development and application of new products in the media and display fields. His background includes development of electronic projection systems, analysis of color reproduction issues in electronic displays, strategic technology sourcing, reviews of advanced electronic projection products, and detailed analysis of compression schemes for digital images. Cowan was instrumental in developing the current mastering processes used in digital cinema, which introduced the use of the digital mastering theater for color and dynamic range adjustment.
Loren Nielsen is ETC's cofounder. At ETC, she focuses on technical issues revolving around achieving theatrical-looking images and business modeling. Her background incorporates more than 15 years of management and marketing of start-up ventures in the technology and entertainment industries. Nielsen has hands-on experience setting up and using high-quality digital viewing environments for feature film applications. She has also served as a producer of high-quality presentations for the demonstration of new technology to the entertainment community, and she is active in the development of business models for digital cinema.
