Advanced Digital Colour Theory

Taught by Charles Poynton

Course Number:
DCT303
Software Version:
 
Original Run Date:
January 2014 
Duration:
8 hours 58 minutes 
pipeline
Colour image coding is at the heart of digital imaging; it is a critical element of high-quality digital cinema and HD production. In this course we follow colour from its spectral form in front of the camera into acquisition where colour is reduced to 3 components, then through the post-production, grading, finishing, approval, mastering, distribution, and display/exhibition stages. We discuss the colour transforms and colourspaces used at all of these stages. We discuss colour aspects of subsampling and compression (H.264, as in Canon cameras, ProRes, as in Alexa, and JPEG2000, as used in DCPs).

We'll present lots of diagrams, graphs, and equations using math at intermediate level; we hope that you're quite familiar with y = mx + B, and somewhat familiar with matrices, power functions, and logarithms. (WARNING: The presenter is Canadian; colour will be spelled with a "u".)
 
Advanced Digital Colour Theory
Watch our overview of the course

Class Listing

Class 1

Technical introduction to colour science: We introduce the spectral power distributions of physics using Newton's drawing from the year 1666. We introduce the CIE standard observer, and the tristimulus values ("tristimuli") that result from applying those weighting functions to SPDs. We derive RGB values, and demonstrate RGB colour mixtures. We review the CMY coding of motion picture film, and review the densitometric coding of classic CPD/DPX systems.

Class 2

Perceptually uniform coding: We define the concept of linearity, and explain why it is important in modern imaging systems. We define lightness. We introduce power functions and logarithms. We discuss the code-100 problem. We ask, what bit depth is necessary to represent reality? What bit depth is required to satisfy human vision?

Class 3

Lin, log, and quasilog coding: There are two kinds of "linear" acquisition and two kinds of "log" acquisition; we elaborate. We outline various proprietary log formats. We discuss theoretical and practical aspects of acquisition coding.

Class 4

Colour transforms: There are many kinds of colour transforms: affine transforms (3x3 linear matrix, in LMS, XYZ, RGB colourspaces, and also in Y'CBCR spaces), projective transforms transforms (used in CIE xy), and various nonlinear transforms (including CIE LAB). We discuss 3D LUTs, gamut mapping, and ICC colour management systems.

Class 5

The ACES workflow, colourspaces, and transforms: Scene-referred linear-light image coding is fast becoming the standard way to do high-end production. We describe the ACES and OCES colourspaces, and the ACES colour transforms: IDT, LMT, RRT, RDT, and ODT.

Class 6

We explain the theoretical and practical necessity of a camera 3x3 matrix. We describe how to build a transform from real camera RGB (device space) to the ACES interchange colourspace - that is, we explain how to build an ACES input device transform, IDT. We also discuss why IDTs have illumination dependencies.

Class 7

Video colour coding: We present a systematic overview of gamma, including the DCI standard for digital cinema (2.6), the BT.1886 standard for HD (2.4) and the sRGB standard for desktop computing (2.2). We review chroma subsampling (4:4:4, 4:2:2, 4:2:0), and discuss its effects on noise, resolution, and bit depth.

Class 8

Distribution: We describe in detail the DCI P3 image coding used for reference D-cinema projectors, and the XYZ coding for DCPs. We outline compression issues associated with ProRes, H.264, and JPEG2000.

Class 9

HD, 2K, 4K ("UHD"), 8K: We detail current and future image formats, and we discuss the underlying theory of spatial sampling, resampling, and oversampling.

Class 10

Emergent display systems: Wide gamut colour is emerging rapidly. We discuss LED backlights on LCD displays; OLED displays; laser displays; spatially modulated backlights and high dynamic range displays. We also discuss the open issue of observer metamerism in emergent displays.