So in March 2002, seven studios—Disney, Fox, MGM, Paramount, Sony Pictures Entertainment, Universal, and Warner Brothers Studios—established Digital Cinema Initiatives, in Los Angeles, to create a specification for digital cinema. They first set a quality threshold: the image resolution had to be, at a minimum, 2048 by 1080 pixels, a resolution loosely called 2K. The systems also had to be upgradable to double that resolution, called 4K. The consortium specified that the systems had to produce essentially the same range of colors as film does, with the future potential to include all colors visible to the human eye.

The group then began looking at how equipment manufacturers could put together a system that would provide that high-quality picture using a minimum of protected intellectual property. The studios knew that the fewer proprietary technologies they chose, the more widely and inexpensively they could implement the systems.

They ended up specifying a video compression technology and recommending uncompressed sound. But they didn’t recommend a specific projection technology, though the industry seems, for now, to have settled on Digital Light Processing (DLP), a micromirror system developed by Texas Instruments. Meanwhile, an efficient method of getting the digital files from the studios to the theaters is still evolving.

The picture, of course, is key. And several things conspire against its quality in conventional film prints. After just a dozen showings, dirt, grease, and scratches visibly degrade the image. What’s more, copying a film print through several generations, which is what the film labs do to generate the immense number of copies needed for distribution, also reduces image quality, in the same way that making a photocopy of a photocopy does.

Starting out, a digital picture with its image clarity and range of color tones and a pristine film print of a movie displayed on a well-maintained film projection system are equal. But the digital version is made with the exact images approved by the director or the studio, and it maintains that quality through an indefinite number of showings and copies.

Today, movies may still be shot on film and then digitized. The digital files typically used in movie production to capture, store, and edit movies after they are shot on 35-mm film are massive, as large as 6000 terabytes. The final uncompressed movie files are a few terabytes. Yet even these files would be too expensive for studios and theaters to store, ship, and handle. 

Obviously, some sort of compression was needed to bring costs down. But the movie industry widely recognized that if digital cinema didn’t start out with quality that was as good as 35-mm film, it was doomed. Selecting a compression technology that would enable digital movies to be packed down to a reasonable size and without any visible loss of quality required Hollywood’s most discriminating observers to do a lot of testing. These “golden eyes” included cinematographers, movie directors, theater ­owners, and studio executives—all people who spend much of their professional careers examining the minute details of images, such as color, contrast, and even the tiniest artifacts that might somehow render an image less realistic.

These cinema experts converged in 2002 on the Hollywood Pacific Theater, a grand old movie palace taken over by the Entertainment Technology Center at the University of Southern California and turned into the industry’s Digital Cinema Laboratory. After replacing the old 35-mm projection systems with the best film projectors available, the group invited digital technology vendors to set up test equipment and asked ­providers of compression technologies to face off against one another. And the games began.

An important technology contender was MPEG-2, the compression system created in 1994 and now ubiquitously used around the world for television, DVD, and Internet video. The problem with MPEG-2, however, is motion artifacts—the appearance of discontinuity or jerkiness in action scenes, particularly ones involving speeding cars or fire.

These motion artifacts appear because MPEG-2 uses temporal compression. The technique essentially encodes only the differences between frames, so, after the initial frame in a scene, the digital files typically need to add very little information for subsequent frames. But in scenes with a lot of action, many changes occur between frames, and the processor that decodes the compressed data cannot keep up, making movement on the screen appear jerky or displaying chunks of the picture as single-color blocks. Motion artifacts are rarely noticeable on a television screen but are all too apparent when magnified on a large screen.

The specification developers also considered the cost of implementation. Companies offered a variety of compression schemes, but many had costly licensing requirements or restricted manufacturing to single sources of supply. The industry wanted quality, but it also wanted an open and competitive market. The group finally settled on JPEG2000 as providing the best possible image quality with the least-encumbered intellectual property.

The original JPEG format (for Joint Photographic Experts Group), established in 1986, is ubiquitous in consumer digital cameras. JPEG is the popular name for ISO 10918-1, a standard created by the combined efforts of many image-processing experts from industry and academia. The format compresses large image files into manageable sizes through so-called lossy compression, a technique that discards some information. When done well, the algorithms discard mostly information that is unimportant to the human eye or to the human brain as it processes signals from the eye.

JPEG2000, first published in 2003, updates that classic technology. It compresses whole frames as if each were a separate picture, which in fact each is. But instead of using the original JPEG algorithms, which analyzed each image and threw away the least important data, JPEG2000 uses “wavelets.”

In this technique, Fourier analysis transforms the image into a set of sine waves with different frequencies and amplitudes. The computer doing the compression then maps the sine waves against a stored set of sine waves—the wavelets—to determine which members of the stored set best represent the image data. The compression program contains mathematical formulas to define each of these stored wavelets and records the image as a set of these formulas. When a computer later decompresses the image data for viewing, it does the math to recreate the original sine waves.