1. ECE 638: Principles of Digital Color Imaging Systems Lecture 3: Trichromatic theory of color

2. Synopsis l Continue development of mathematical framework for trichromatic theory

3. Review: foundations for trichromatic theory l Color matching experiment l Grassman’s laws for color matching l Spectral models for color l Surface-illuminant interaction model

4. Review: color matching experiment

5. Review: spectral representation of color Key Result:

6. Development of trichromatic theory l Sensor model l Reinterpretation of conditions for color match l Metamerism l Linearity of sensor model l Response to monochromatic stimuli l Chromaticity diagram

7. Trichromatic sensor model l are spectral response functions that characterize the sensor

8. Trichromatic sensor model (cont.) l With suitable choices for the spectral response functions, this model applies to the HVS, as well as color capture devices, such as digital cameras and scanners. l The 3-tuple response vector represents the system output for color capture devices and an internal signal for the HVS. The subscript ‘S’ refers to the particular stimulus, in this case

9. Trichromatic sensor model for HVS l When used to model the HVS, it is common to use the following terminology: –Red (R) Long (L) –Green (G) Medium (M) –Blue (B) Short (S) l In this case, the sensor response vector is called the tristimulus vector

10. Color matching condition l Two stimuli will match to a human viewer if and only if they have identical tristimulus (vector) values, i.e. where and and are the spectral power distributions corresponding to the stimuli and

11. Example 1 l Ideal block sensor l Find response to two different stimuli

12. Example 1 (cont.) l Sensor responses to both stimuli are identical l What are the implications of this fact? –their spectra are quite different –but sensor cannot distinguish between them

13. Metamerism l Stimuli and are said to be metameric with respect to the sensor if they elicit identical responses l Metamerism is both a blessing and a curse for color imaging systems designers

14. The blessing l Consider for the moment if there were no metamerism –i.e. every pair of distinct spectral stimuli and yielded distinct responses and l What are the implications for color imaging systems design? Faithful color reproduction would require exact replication of the spectrum of the original object or scene l Can a trichromatic sensor system achieve this kind of performance? No, the space of all spectral distributions is infinite- dimensional. The space of all trichromatic sensor responses is 3-D

15. The curse l Consider two different sensors l and two different stimuli l Suppose that l But

16. What are the implications of this situation… l If sensor is that of the HVS, whereas is that of a digital camera? l The camera cannot faithfully reproduce color as seen by the human viewer. l Does this mean that the spectral response functions of a digital camera must be identical to those of the HVS? l We will address this issue later. l How about if sensor is that of the HVS under one illuminant, whereas is that of the HVS under another illuminant?

17. Role of illuminant l Consider the response of sensor to an object with reflectance observed under illuminant l The stimulus is then l and the response is given by

18. Role of illuminant (cont.) l By grouping the illuminant spectral density with the sensor response functions, we obtain a new effective sensor with response to the surface with reflectance viewed under an equal energy illuminant where

19. Now consider again the question posed earlier l We have one sensor (the HVS) l two different surfaces with reflectances and l two different illuminants and l Under the first illuminant, the two surfaces appear identical, i.e. l Under the second illuminant, they do not; so l This situation can be a serious problem for color process control and color imaging systems

20. Examples l Color matching of different materials –fabric and buttons for clothing –instrument components and dashboard in automobiles –natural teeth and restorations in dentistry –we may get a match under indoor fluorescent lighting, but not under direct sunlight l Digital photography –The same objects photographed under daylight and indoor tungsten light –The human observer adapts to reduce the color shift caused by the changing illuminant – the camera does not

21. Linearity of sensor response l Recall again our trichromatic sensor model l It follows directly from the linearity of the integral, that the trichromatic system is linear

22. Response to monochromatic stimuli l From the linearity of the sensor system, it follows that the response to any stimulus can be expressed in terms of the response to monochromatic stimuli l is the “impulse response” of the system

23. Principle of superposition l For an arbitrary stimulus, we can write l Each component of this stimulus generates a scaled version of the impulse response

24. Principle of superposition (cont.) l We sum responses to stimuli at each wavelength to get total response to the stimulus l Note that sensor systems do not obey a principle corresponding to shift-invariance.

25. Geometric interpretation of the sensor response l The sensor response is a 3-tuple; so the response to each stimulus may be viewed as a point in Euclidean three space that we call the sensor space l Note that the axes need not be perpendicular for this representation to be meaningful

26. Effect of scaling the stimulus l If we scale the stimulus by a constant, the response will scale by the same amount l In the sensor space, the response point will shift along a straight line as we vary

27. Sensor chromaticity diagram l Since scaling the stimulus by a constant does not modify the relative spectral power at each wavelength, we intuitively expect that such changes will only affect the lightness, and not the saturation or hue of the resulting color l Thus, if we are only interested in studying how the saturation and hue vary with the spectral shape of the stimulus, it is sufficient to consider the point of intersection of the sensor vector with a fixed plane in the sensor space

28. Plane of equal lightness (brightness) l Intuitively, we also expect that all points in the sensor space for which for some constant, will appear equally light or bright (in reality, this is only approximately true) l Thus we choose the plane for our sensor chromaticity diagram

29. Orthogonal vs. non-orthogonal axes l When the axes are orthogonal, the triangle is equilateral with sides l It is of interest to choose a particular set of non- orthogonal axes for which we have an equilateral triangle with all perpendiculars having length

30. Chromaticity diagram l In this case, the lengths of the perpendiculars illustrated below can be shown to satisfy

31. Shape of sensor response functions l Under our trichromatic model, the sensor behavior is entirely determined by the shape of its response functions l As we develop a more complete vector space view of the sensor system, we will gain a better understanding of this relationship l For now, we examine the sensor chromaticity diagram to gain insight