Re-defining the colour rendering index

1
Re-definingthe colour rendering index
Peter Bodrogi Stefan Brückner Tran Quoc Khanh
2
New colour rendering index our point of view

• In practice, a rank order scale may be more
  usable than a continuous scale
• It is easy for the user to understand very good,
  good, acceptable, not acceptable and very bad
• What is the visual difference between Ra78 and
  90? And between Ra78 and 82?
• New CIECAM02-based colour spaces like
  CIECAM02-LCD (Luo et al, CRA, 2006)
• Define the category limits for very good, good,
  acceptable, not acceptable and very bad in a new
  visual experiment in terms of CIECAM02-LCD
• Computed categories for a set of test colour
  samples, e.g. for a new white LED light source
• skin very good banana good cucumber
  acceptable etc.
• Compute from the above rating vector a single
  quantity like its median and then the new CRI
  like 6-median(rating vector)

3
New questionnaire
Variable DEvis
Variable R
Variable P
4
New definition of the colour rendering index
based on computed ratings
13.7
9.5
7.25
5.5
5
New definition of the colour rendering
indexhistograms of the computed ratings
6
New definition of the colour rendering index1
median scale (coarse)
CRITU_DA_1 (100/5) (6 - Median RCi i117)
7
Comparison of CRITU_DA_i with Ra (1-8) and Ra
(1-14) and with CQS7.1
8
New colour rendering index our point of view

• Separate indices for the users different tasks
• Naturalness or colour fidelity the true
  appearance of simple standalone surface colours
• as under A or D65 or from long-term
  colour memory
• Saturation enhancement can be in a separate index
  (preference)
• CPI for preference or flattery CDI small
  colour differences
• HRI for harmony rendering (is the appearance of a
  combination of colours aesthetic?)
• Visual clarity, large colour differences
• Acceptability of realistic scenes with many
  colours
• Every user can select an index appropriate for
  the task, e.g. fidelity can be important for
  textile designers
• Or CDI for electricians working with wires
• Weighting of the indices

9
New colour rendering index our point of
viewAdditional index for inter-observer
variability
10
New colour rendering index future tasks

• CATs mostly collected under two light sources
  (D65 and A)
• Combinations of surfaces, textures and shapes can
  evoke higher order neural mechanisms affecting
  colour constancy e.g. memory, language, or
  object recognition
• Compare the colour constancy index (e.g. the
  Brunswick ratio) with the colour rendering index
• Effects of local contrast, global contrast, and
  image content on chromatic adaptation
• Relational colour constancy spatial cone
  excitation ratios in a multi-colour-surface scene
  remain (relatively) constant under changing
  illuminations but this may not hold for novel
  artificial light SPDs
• Chromatic textures in natural scenes the
  discrimination of texture stimuli is not the same
  as the discrimination of uniform colour patches
  under the test and reference illuminants
• Contextual factors higher order processing
  involving long term colour memory for familiar
  objects (e.g. human skin tone). Pictorial
  (hyperspectral) test images are needed instead of
  standalone patches.

11
New colour rendering index future tasks

• The colour rendering index can be improved by
  advancing two lines of research
• 1. computational colour constancy
• 2. colour image difference metric
• Our planned experiments
• Double-chamber viewing booth experiments
  currently underway
• Aim 12 observers x 2 CCTs (2900K and 5000K) x 5
  light sources x 20 test colour samples
• Validate the scale of the new CRI_Darmstadt (very
  good-very bad)
• Develop the inter-observer variability index
• Tabletop (still life) with several coloured
  surfaces (textures, objects) and a test room with
  immersion in the visual environment
  (acceptability study) with several conventional
  and new light sources

12
(No Transcript)