Computer Systems for Specifying Visualizing and Communicating. Colour. David Oulton Manchester University May 2009. Introduction. A review is presented of ...
Computer Systems for Specifying Visualizing and Communicating Colour David Oulton Manchester University May 2009 Introduction A review is presented of a range of techniques and modeling methods that allow colour to be accurately specified, visualized and communicated. The best available starting point for the relevant computer models is the CIE (Commission Internationale de l-Eclairage) Standard Observer and Standard Illuminant model. This internationally standardized model generates CIE XYZ and L*a*b* colour co-ordinates. These successfully predict visual colour matches, and quantify both visual identity and colour difference. R.S. Berns gives a full account of the CIE system in reference [1]. Traditionally the eventual colour of a textile product is specified using physical samples that must be matched either visually or with the aid of colour measuring instruments. The advent of computer generated colour images has opened up significant possibilities for generating and communicating ‘virtual samples’ [2,3]. Apart from saving both time and money the resulting visualizations enable global electronic collaboration and improve the flexibility and speed of response to the market. Using recently developed colour calibration and product visualization techniques the stakeholders in colour related products can establish colour communication and colour management networks. The objective is threefold:1) To manage the specification, visualization and production of coloured products. 2) To optimize colour choice. 3) To reproduce the original design inspiration rapidly and faithfully.
The Colour Optimization Network The colour optimization network links the designers and manufacturers to the managers and policy makers at all stages of product development, production and sale.
COLOUR DECISION INTERFACES Design Input
Marketing Input
Technical Performance Control
Range Policy
Complex Coloured Product Suppliers
Packaging and Display
Customer Interaction and Response
Clearly all stakeholders will seek maximum benefit from shared information and they should also benefit from any customer feedback generated by the retailer. The network should also enable rapid response supply within a globally distributed supplier-buyer network. The objective is a cohesive team with many distinct interests. The input from the designers and the descriptions used by them are primarily with product concept, colour and colour co-ordination, and a key objective of the colour optimization network is to ensure that the original design inspiration is realized in the eventual product. The management team will be concerned with product range content, exact specification of colour, product practicality, cost, availability, continuity and co-ordination. The marketing personnel are concerned with product and brand identity, cost, target markets, fashion and customer feedback. The production team are also concerned with establishing product practicality, but have specific technical concerns over the required colour specification, dye choice, colour fastness, production management and delivery. The ultimate retailer will be concerned with current fashion, seasonal range management, store presentation, sales monitoring, supply chain issues and customer feedback. Each of the above stake holders has distinct concerns and distinct colour specification requirements. The key aims in a colour optimization network are therefore to:1. Present clear easy to understand accurate and timely information to the relevant people. 2. Relate together the creative, managerial and technical aspects of colour, as attributes of each product or service. 3. Generate meaningful colour descriptions and specifications in all communication links.
Virtual Products Virtual products based on high quality images and accurate colour specifications can be used to replace physical coloured samples during product development. Significant savings in both time and costs, leading to reduced lead times have been demonstrated [3] using such virtual products. Fully developed colour networking systems such as CHROMASHARE tm [4] use a series of models and specification transformations to aid communication. These include:1. A direct link to recipe libraries and dye recipe formulation 2. Comprehensive tools for changing both the imaged colour and texture as independent variables of colour appearance. 3. A model of the effect of surface texture on the perceived colour in photo-quality images. 4. Precise on-screen colour calibration.
The workflow for colour idea development from inspiration ... to concept colour Manipulating colour in images, extracting colour from them, managing colour palettes and sharing colour ideas
Appearance preview Colour Library search
Evaluating colour within textures and garment styles
Fast, flexible colour searches to find the right shade
© Chromashare Ltd.
The workflow from concept ... to product Scaling colour concepts up to production level, supply chain communication,
Colour Library search Fast, flexible colour searches to find the right dyeing
product optimization
Match appraisal Evaluating colour within textures and garment styles
Supplier - buyer workflow system for colour sampling.
Submission acceptance management Colour quality and product continuity evaluation
© Chromashare Ltd.
SPECIFYING COLOUR BY VECTOR SUM The cause of colour has N defining elements or dimensions quantified by wavelength RGB Response weightings
x
Mapping by vector sum
Description of the colour using weighted coordinates
S .R .x X
S
Illuminant SPD
S .R . y Y
y
R z Surface colour Reflectance
S .R .z Z
The CIE X,Y,Z colour co-ordinate system successfully predicts colour identity and visual colour matches because it models both the colour sensation of a standardized observer, and the effects caused by changing the lighting conditions [1].
SPECIFYING COLOURS by their X Y Z CO-ORDINATES on a COMPUTER COMPUTER RGB to CIE XYZ CO-ORDINATE CONVERSION by FORWARD AND INVERSE MAPPING USING LINEARIZATION FUNCTIONS and MARTRIX CROSS DEPENDENCY [2 , 3] LINEARIZING the COLOUR SPECIFICATION
Specification
Y =fn(X)
DRIVE VALUE
Drive value Y f n X
R f1 R G f 2 G B f 3 B
LINEARIZING the DRIVE SPECIFICATION
R f1R G f 2 G B f 3B
SPECIFICATION
LINEAR CROSS DEPENDENCY
X Y Z R G B
R * G B
M
M
1
X * Y Z
Image Segmentation and Colour Manipulation Each image represents a combination visually distinct objects that can be separated logically via their CIE L*, C* H0 colour co-ordinates.
Colour-Set membership in a Vector Space L
Each of several thousand colour definitions is represented, and related by a colour difference L, C, H to a base-colour.
Base (Intrinsic) Colour
Many pixels are linked to each colour definition. Pixel Colours Form a Variable Density Cloud, round the base-colour definition.
C H
Set of Pixels
The image can be segmented into a set of distinct objects by reference to their colour, and the surface texture appearance of each object is analysed by reference to the Lightness, Chroma and Hue (L C H) co-ordinates of each image pixel that represents part of the object [5].
Vector Analysis of Colour-Set Membership ABSTRACTED IMAGE OBJECT
ABSTRACTED TEXTURE CLASSIFIED BY COLOUR
COLOUR SET CONTENT ANALYSED BY L, C AND H
COLOUR SET
L,C, AND H DISTRIBUTIONS
Taking into account the effects of lightness and shade, the colour set members have quite a wide range of co-ordinates
Pixel Frequency
LIGHTNESS DISTRIBUTION 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0
8
17
20
23
25
27
29
31
33
35
37
40
42
45
52
Pixel population distributions For a medium depth medium chroma yarn winding, judged by observers to have a single intrinsic colour.
L value
CHROMA DISTRIBUTION
Pixel Frequency
7000 6000 5000 4000 3000 2000 1000 0
5
14
16
17
19
20
22
23
24
26
27
29
31
33
C Value
HUE DISTRIBUTION Pixel Frequency
12000 10000 8000 6000 4000 2000 0
66
114
120
125
131
137
143
148
154
H Value
Note that the Lightness and Chroma scales each have about 100 subdivisions and in order to represent the texture, the range of pixel co-ordinates that specify L and C light and shade is substantial. By contrast, within the 0 – 360 range of H (Hue angle) the H values are much more tightly grouped. Hue angle is thus a useful object discrimination tool.
Vector Analysis of Colour-Set Membership The Colour-set measurements MDDL, MDDC, and MDDH, define the combined effect of all the individual shade variations across the texture.
COLOUR SET M.D.D MEAN DIRECTIONAL DEVIATION MDD And Mean Vector Displacement MVD Tex
MDDC
MDDL
MDDH
M.V.D. MEAN VECTOR DISPLACEMENT M.V.D.tex =
[ (L), (C), (H)] n
n
n
Colour-Set Analysis
Hue Distribution is a key analysis tool for separating objects.
The cylindrical / polar co-ordinates C* (radius) and H0 (hue angle) are derived from CIE L*a* b* coordinates as described by Berns [1], and Hue angle difference is optimal for differentiating objects.
ORIGINAL IMAGE
SAME IMAGE MODIFIED FABRIC COLOUR
The computer generated logical separation of the satin fabric is almost complete, and it can be recoloured on demand. When colour sets are moved as a group in CIE LCH colour space, the variations of light and shade (i.e. in the L* dimension) are preserved.
Intrinsic Colour and Colour Appearance
•
Intrinsic colour is defined as a central property of the
•
Colour Appearance is produced by applying the intrinsic
colorants present (product colour specification). colour to a given Textured Material.
MVD1 Intrinsic Colour
Object Colour-Set (texture 1)
Colour Appearance (1)
(of The Recipe)
MVD2
Object Colour-Set (texture 2)
Colour Appearance (2)
Colour appearance variation
MDDL (CIEL*a*b* units) from winding Tuft = -27.4, Knit = 11.8 Colour Appearance variation : One Yarn, Three Textures
One lipstick colour, four alternative colour specified skin tones. Conclusions Both technically significant and commercially important conclusions may be drawn from the use of computers to quantify, visualize, manipulate and communicate colour. In terms of technical capability, it is possible to:1. Segment high resolution photographs into sets of logically distinct objects that can be manipulated separately at will on the computer screen. 2. Quantify the effect of surface texture on colour appearance and visualize the relevant colour differences. 3. Calibrate the colour reproduction of the computer screen, so that for any given colour, it is a visual match to any physical sample with the same CIE co-ordinates. 4. Bring together all the data that is logically related to a specific product or batch. Thus, colour names, batch numbers, spectral reflectance data and general product management data become more manageable over a multi-site globally distributed network. In commercial terms, it becomes possible to:1. Use ‘virtual products’ to reduce the need for physical samples during product and colour range development. 2. Use electronic networking to make both technical and managerial communication faster and easier. 3. Establish much closer supplier-buyer collaboration. 4. Gain competitive advantage in terms of flexibility, cost and speed of response to the market.
References 1. Billmeyer and Saltzman’s “Principles of Color Technology” 3rd Edition; Ed. R.S Berns, Ch 2 pp 31 – 75. Publ. John Wiley & Sons New York. 2. "Control of Colour, Using Measurement and Feedback"; D.P. Oulton and I. Porat, J.Text.Inst. 1992, vol. 83, No.3, p 453 et seq. Published by the Textile Institite Mancheaster UK. 3. “Building a Precision Colour Imaging System”; D.P.Oulton, J.Boston, & R.Walsby, Proc. IS&T/SID 4th Int. Imaging Conf. Scottsdale Arizona, pp14-19 Nov 1996. 4. Chromashare http://www.chromashare.com/ 5. “Quantifying the Effect of Texture on Colour Appearance” DP Oulton. AW Bowen and E. Peterman; Colour Science Vol 3: Colour Physics pp 12 – 21; Ed. A Gilchrist and JH Nobbs. Pbl. by University of Leeds Print Services, Leeds LS2 9JT, UK.