Digital Intermediate

Any discussion concerning Digital Intermediate begins with an understanding of the

processes involved in today’s filmmaking⎯film origination, digital image processing,
and film distribution. Below is a chart of the “Motion Picture Imaging Chain” that
represents an overview of the three parts of the Process⎯Image Capture, Image
Processing, and Image Distribution. Several years ago, Kodak pioneered digital post-
production with the advent of KODAK CINEON products. Although these products are
no longer available, they were pioneers of this process and formed the infrastructure on
which the industry continues to build. Optical effects have almost entirely been replaced
by digital workflows, as have special effects.

Motion Picture Imaging Chain

Digital Intermediate is about using digital imaging science with film-originated images.
Every single image of film is digitized and imaging algorithms applied for special effects,
color correction, contrast manipulation, restoration, adding computer-generated images,
dustbusting, wire and boom removal, speed change effects, certain creative looks, and
look changes such as sharpening, degraining, or color timing.

Using the Digital Intermediate process, or DI, provides access to not only digital imaging
science, but also to a very highly capable digital infrastructure (computers and
workstations, high-quality displays, networks, data storage). It also requires high

1
resolution, high throughput equipment for film scanning and recording. Data storage
such as storage area networks (SANS), or network-attached storage (NAS), can allow
different flexibilities and capabilities for manipulating, storage, and processing of images.
There are many hardware and software vendors that have products that take advantage of
these systems.

The story of Digital Intermediate begins in the 1990s. Kodak introduced the term with
the introduction of their Cineon system, which included the Genesis digital film scanner
(capable of scanning up to 4K resolution), the Lightning laser film recorder (also 4K
recording), and Cineon software (which was supported by the Kodak digital imaging
system with an object-oriented workflow or flow graph that served as a model for the
industry). Originally, “Digital Intermediate” was a delivery element. It included
scanning at the highest resolutions available, creation of digital data files, and output to
film for theatrical distribution⎯Film In, Digital Manipulation, Film Out.

“Pleasantville” was the first movie to utilize a significant amount of negative

digitalization, but “Oh Brother, Where Art Thou?” was the first completely
digitized⎯Digital Intermediate⎯feature film.

From the beginning, DI has faced the major obstacles of productivity and cost. Initially,
DI was a very expensive and slow post-production process. Recording images and data
file transfers in a facility remain the chokepoints in the DI process. However, the
continuous development of key technologies such as real-time high-resolution data
scanners, higher speed recorders, faster computers and workstations, very high capacity,
and affordable storage are making DI more economically viable today. More than 400
movies have gone through this process over the past few years and today, DI is a reality
for more and more customers. It is becoming the “norm” just as nonlinear editing and
digital sound have.

Digital Intermediate means post-producing a movie in a digital format. The digital

formats can vary from post-production facility to facility. Once the Digital Master is
ready, an intermediate is recorded back to film, and the final prints produced in the usual
way. However, this same “digital master” can also be converted to DVD or TV
broadcast formats, or as a movie that can be projected in a Digital Cinema theater. Each
viewing medium has its own color gamut and must be graded on the intended display
device for preserving the creative content of the director and cinematographers.

DI today relates to any long-form project that is entirely post-produced digitally. The
Image source input can be from color negative film and/or digital cinematography
cameras (RGB data, not video). The final result is a Digital Source Master from which
all film and video deliverables can be produced. For film prints, a Digital Negative is
produced for making either an IP/DN or Direct prints for theatrical release.

For Filmmakers, DI offers the following benefits:

• Flexibility and immediacy
• Multiple and different versions of content, trailers, and teasers
2
• Localized version of language or subtitles
• Full color correction (i.e., primary and secondary)
• Full editorial capabilities
• Large screen displays for evaluating interactive digital image process
• Less costly last-minute changes
• High image quality of optical effects (fades/dissolves, composites, etc.)
• Image repair
• Text capabilities (titles, subtitles, and credits)
• As DI processes continue to improve, it will result in image quality improvements in
the 137,000 film cinemas around the world

However, there are many disadvantages as well:

• Too many creative opportunities and decision makers, which can lead to a time-
consuming process (“fix-it-in-post syndrome”)
• Expensive hardware, software, and networks are required
• Equipment obsolescence
• No long-term digital archiving medium except on film
• High level of expertise and complementary skills required
• Low-speed image processing (rendering) and film recording

And what DI means depends on whom you ask and where you are in the workflow. As
processes unfold and as digital tools are applied, so do the terms customers use.

In today’s digital environment, there are a number of options that can be used throughout
the entire DI process. However, with each option there is the potential that the integrity
of the film images will be compromised. How Images are input is the first of many
choices customers need to make as their project enters the DI workflow.

3
 What do we call the output of a DI Workflow?

Scanned Digital
Digital
Scanned “Optical
Original
OriginalCamera
Camera “OpticalEffects”
Effects”
Negative (e.g.
(e.g.Dissolves,
Dissolves,
Negative Flip/Flops…)
Flip/Flops…)

Digital Capture Digital Digital Scene-to-Scene

Digital Capture Digital Dirt Digital Digital Scene-to-Scene
Data (raw) or Digital Dirt Visual Color Timing Digital
Data (raw) or Conform to Cleanup Visual Color Timing Digital
Video Conform to Cleanup Effects Secondary
SecondaryColor
ColorCorrect
Correct Intermediate
Video EDL @@2K Effects Intermediate
EDL 2K (2D, Sharpness
(HD, SD)
(HD, SD) (2D,3D…)
3D…) Sharpness
Noise/Grain Management
Noise/Grain Management

Output
CGI
CGI Digital
Files
Files Digital“Squeeze”
“Squeeze”
Super
Super35
35––Anamorphic
Anamorphic
Super 16 to 35
Super 16 to 35

Input Manipulate

Besides film, there are many types of images captured electronically. There are also
computer-generated images (CGI), which may become part of the workflow. Both of
these are generally characterized as linear data. How this data is stored and later,
manipulated, are additional, important choices that customers make. Each digitally
captured or (CGI) is used in a different workflow than film-generated and scanned
images.

In the middle part of the workflow is the greatest potential for data to be compromised.
There are many different types of digital tools at work here⎯hardware and software:
displays, processors, data networks, etc. As data moves from one step to the next,
decisions concerning data spaces and color management are key decision points for
preventing the compromise of quality and integrity of the images.

A Closer Look at the DI Process

The typical flow can be analyzed in terms of the following three major steps:
• Input
• Manipulate
• Output

4
 “DI”, A Closer Look

Scanned Digital Film

Scanned Digital Film
Original Camera “Optical Effects” Negative
Original Camera “Optical Effects” Negative
Negative (e.g. Dissolves, Master
Negative (e.g. Dissolves, Master
Flip/Flops…)
Flip/Flops…)

Digital Scene-to-Scene Digital

Digital Capture Digital Digital Scene-to-Scene Digital
Digital Capture Digital Dirt Digital Color Timing Distribution
Data Digital Dirt Distribution
Data(raw)
(raw)oror Conform to Cleanup
Visual
Visual Color Timing
Secondary Color Correct Master:
Video Conform to Cleanup Effects Secondary Color Correct Master:
Video EDL @ 2K Effects Sharpness Video
(HD, SD EDL @ 2K (2D, 3D…) Sharpness Video
(HD, SD (2D, 3D…) Noise/Grain Management (HD,SD,DVD)
Noise/Grain Management (HD,SD,DVD)

Digital
Digital
Distribution
CGI Distribution
CGI Master:
Files Digital “Squeeze” Master:
Files Digital “Squeeze” E-Cinema,
Super 35 – Anamorphic E-Cinema,
Super 35 – Anamorphic D-Cinema
D-Cinema
Super 16 to 35
Super 16 to 35

Input Manipulate Output

The Input is a data file. “Images” are produced from high-resolution (>2K) scanned
film, lower resolution DVCAM, HDCAM, HD, standard def video, computer-generated
RGB images, or from a variety of Digital Cameras with varying resolution algorithms.

Manipulate consists of a whole range of transforms to which the images will be

subjected: color correction, look management, special effects, editing, conforming,
formatting, and fades/dissolves, etc.

And Output is the form in which the movie will be distributed. That includes film,
digital projection, HD, DVD, and standard definition video.

Every image at every step in the DI Workflow is characterized by two attributes:

resolution and color space.

Resolution is measured spatially, in numbers of lines in both the X and Y axes. Color
Space is measured in bit depth, in how many units of color and luminance present in a
digital image. Both of these attributes affect the integrity and quality of the Digital
Intermediate Master, or Final Output.

Below is a useful diagram for examining the choices available at each step of the DI
Workflow:

5
 “DI”, A Closer Look

Film @ 16b Lin Film/D-Cinema

Up to 6K 4K | 4K
| |
Digital Capture 10b Log |
3K
“Raw” Data | I Film/E-Cinema
(up to 4K) 2K 14b Lin 2K
| |
| |
HD Video HD 4:4:4 HD Video
(1920x1080) | 12b Lin
|
|
HD Video
HD 4:2:2 | SD Video
|
(1280x720) 10b Lin |
| SD 4:4:4
SD | | DVD
(720x480) SD 4:2:2 8b Lin |
Other

Input Manipulate Output

Input files can originate from the lowest resolution and color space of digital video to HD
video and up to the maximum resolution and color bit depth of scanned film. Likewise,
Data manipulation can occur at various resolutions and color spaces—from Standard
Definition video up to 12-bit logarithmic data: the highest resolution and bit depth of
scanned film. And finally, Output follows the same pattern, from the lowest resolution
and color space of video up to that of film recorded at 4K.

The integrity of the Input files and how they are manipulated will affect the resulting
quality of the Output.

Obviously, footage recorded on a DV camera can be recorded to film at 4K, and film
scanned at 6K can be output to a DVD; however, choices must be made at each step in
the DI Workflow that affect the resulting integrity of the Digital Intermediate Master
(output). However, when the input is of higher resolution (e.g., scanned film), going
downward in resolution and color space is always feasible. The reverse⎯going upward
in resolution and color space from video to film⎯is not as high quality in the end result.
This is referred to as up-sampling and many algorithms are used to do this.

“Video to film” spans from lower end image capture to HD 8:8:8. Keep in mind that an
8:8:8 format results from the Y, Cr, and Cb samplings at every pixel site, and
intermediate values are computed by interpolation to add precision in image transforms
as rotation.

6
In technical terminology, this is not a Digital Intermediate. The input and the image
manipulation steps are of lower quality and may introduce digital artifacts in the output
image.

Video to Film…

FILM @ 10 Bit Logarithmic FILM @

6K, 4K,2K | 4K
| 14 Bit Linear |
FILM @ | FILM @
1920 to 2K 12 Bit Linear 2K
| | |
FILM > HD 4:4:4 HD Video
HD Video | |
| HD 4:2:2 SD Video
HD Video
|
Video to Film| |
DVD
SD 4:4:4
DVcam | |
| SD 4:2:2 Etc…
Etc… Etc…

Input Manipulate Output

Although Digital Video post-production operations are up the quality chain, Digital
Video is not Digital Intermediate either.

7
 Digital Video to Film…

FILM @ 10 Bit Logarithmic FILM @

6K, 4K,2K | 4K
| 14 Bit Linear |
FILM @ | FILM @
1920 to 2K 12 Bit Linear 2K
| | |
FILM > Digital Video
HD 4:4:4 HD Video
HD Video | |
| HD 4:2:2 SD Video
HD Video | |
| SD 4:4:4 DVD
DVcam | |
| SD 4:2:2 Etc…
Etc… Etc…

Input Manipulate Output

Digital Intermediate applies to film scanned at 2K resolution minimum and digitally
processed with the highest resolution bit depth and format at all times.

Film to Film… High End Digital Intermediate

FILM @ 10 Bit Logarithmic FILM @

6K, 4K,2K
|
Digital Intermediate
|
14 Bit Linear
4K
|
FILM @ | FILM @
1920 to 2K 12 Bit Linear 2K
| | |
FILM > HD 4:4:4 HD Video
HD Video | |
| HD 4:2:2 SD Video
HD Video | |
| SD 4:4:4 DVD
DVcam | |
| SD 4:2:2 Etc…
Etc… Etc…

Input Manipulate Output

8
The three key aspects of Image Quality that are part of the DI process are:

• Image space
• Resolution and sharpness
• Color management and calibration

Image Space

The human perception of color, details, and lightness is based on the utilization of cones
and rods, each having different spectral and absolute sensitivity. The human eye can be
considered as a digital camera with a variable focus reflex lens, gray scale sensors (rods),
individual red, green, and blue sensors (cones) with large size variations, arranged in a
high-frequency pattern in the retina, and a very sophisticated image processing system in
the brain that is totally image adaptive. For example, we see a difference in light from 1
to 2 the same way as 100 to 200. That is considered a logarithmic scale and is the same
way that film sees light.

The human visual system or HVS can be described as a Red, Green, Blue, and
Luminance system working in a log space. This is a chromaticity diagram that depicts
the different color gamuts of different systems.

9
Obviously, many different technologies can represent an image: negative film onto print
film, CCDs, inkjet printers, scanners, recorders, CRT displays, LCD displays, and Micro-
mirrors in some digital cinema projectors. All of these technologies can provide a
representation of an image. However, each of them will use a different image space.
Some will be linear, some will be log based. Some will be RGB, some will be YUV,
some will be color subsampled, and some will not. They will vary in terms of bit depth,
spatial resolution, color coding, etc.

The image space will define how a given technology represents an image. In the diagram
above the “Color Space” for Color Negative Film (ECN), Print Film (ECP), and HD
Video have been overlaid to illustrate the limitations of each.

Resolution and Sharpness

The chart below reports the pixel resolution obtained by scanning 35 mm format film
using sensors capable of capturing from 2K to 8K samples across each frame line.
Higher resolution scanning preserves more of the image detail captured in the film.

Keeping the detail information captured by film throughout the entire DI process ensures
that what the cinematographer has seen and wanted will be actually reproduced. Scanner
technology is improving, and 4K is being more widely used today.

How Many Bits?

Resolution is only one part of the story. Resolution applies to the spatial sampling of the
image. The other part of the digitization process is quantization. The human eye is more
sensitive to equal brightness changes in the shadows than in the highlights. By using
logarithmic sampling such as film density, more intensity levels can be allocated to
represent shadow detail. In order to approach our Human Visual System, a 10-bit log
sampling is often used, as it allows a better representation of the shadow details.

However, some systems work in a linear space where10-bit log is the equivalent of 12- to
13-bit linear. As computer power increases, and as the demand for consistency along the
imaging chain is of paramount importance, more bits will be used in the future.
Eventually, higher bit depth linear may be the choice (32 bit is beyond eye capability).

10
When images are captured with a digital camera having a single sensor and a color filter
array (CFA), each pixel captures image information in only one color (red, green, or
blue). To obtain a fully populated color plane (R,G,B at each pixel site), like film does, it
is necessary to mathematically generate the two missing colors at each pixel site. This
mathematical process can introduce chromatic artifacts that significantly reduce image
quality, particularly when subjects with repeating patterns and motion are captured.
There are digital cameras available that use three separate sensor arrays to capture R,G,B
information at each point in the image plane, but they have complicated relay optics,
require higher bandwidth storage, and are incompatible with 35 mm Cine lenses.

There are also Resolution compromises when it comes to the Aspect ratio of the Image.

11
During image processing of the data, there are many possible areas where the image
integrity can be compromised. Digital Intermediate Workflow is much more complex
than a Video Post-production.

12
Digital Video Post-Production vs. Digital Intermediate
Digital Video Post- Digital Intermediate
Production (DV) (DI)

In a stream like water in a In computer files – generally CINEON

Data
pipe or DPX
Low precision – Generally
High precision - 10 bit log going to 32
Precision some form of 8 to 12 bit
bit FP
linear with gamma
Speed Real time Seconds per frame

Resolution Up to 1920 pixels/line 2048 to 6224 pixels/line

Color sub-
Almost always Never
sampled

Compression Always Never

Dynamic
Supports 10 stops Supports 20 stops
Range

Color Management and Calibration

Consistent image color is needed throughout all operations of the DI workflow.

Color Management is the use of appropriate hardware, software, and methods or
procedures to control and adjust color in an imaging system (“WYSIWYG” concept).
The final output must match the image seen in the Digital Intermediate suite. This
requires that each part of the workflow be calibrated independently. There needs to be
assurance that errors in one part of the workflow are not compensated for by more errors
later in the workflow. Images can be exchanged between post-production houses at any
time, and this requires some sort of common format. For now, the Kodak Cineon file
format, ratified by SMPTE as the DPX file, is the cornerstone of the Digital Intermediate
Workflow.

But even the best color management system will not provide good results if the various
elements of the workflow are not perfectly tuned or calibrated to a standard. This applies
not only to the monitors and digital projectors that are used in the workflow but to
scanners and recorders as well. Every single element of the workflow has to be properly
calibrated. A post-production facility can be internally calibrated, but may not be able to
give a data tape to another facility and them being able to use it appropriately to achieve
the highest quality product. Lab processing is also a consideration when using film
systems and need to be accounted for in an ideal calibrated system.

The same sequence of images will eventually go through various digital steps, such as
contrast adjustment, exposure correction, and special effects. Each step may involve a
different image space. In order to avoid any degradation of these images, a device-
13
independent image space is required. For this reason, the DPX file format is the current
format of choice for the Digital Intermediate Workflow.

After the program has been managed in the workflow to the point of distribution, it may
be sent to a CRT, LCD, or laser film recorder to produce the Digital Negative, Digital
Interpositives, or even a Digital Print. The common practice today is to output images in
2K resolution at relatively slow frame rates. These Digital Negatives are used to produce
either Direct Prints for Release or Interpositive/Duplicate Negatives for Release printing.
As recorder technologies improve, higher resolutions and faster recording rates will
become possible. There are some higher resolution scanners and recorders available on
the market, but the data pipeline then becomes cumbersome and expensive. The very
high resolutions are used on large format and special projects.

For Digital Cinema Distribution, a Digital Cinema Distribution Master (DCDM) is

produced. Currently this Master is formatted according to the Digital Cinema Initiative
specification and SMPTE DC-28 standards that will be used for playback and display.
Piracy and security are also important parts of the digital cinema standards.

For the lower quality outputs, digital masters are made for producing DVDs, HDTV, and
SDTV Broadcast versions, etc.

If the integrity of the Image has been maintained throughout the Digital Intermediate
Workflow, the customer can be assured that every version of the project⎯35 mm film
print to DVD⎯will have the best image quality possible.

14