Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Withouck et al. Vol. 30, No. 6 / June 2013 / J. Opt. Soc. Am.

A 1

1 Brightness perception of unrelated self-luminous colors

2 Martijn Withouck,1,* Kevin A. G. Smet,1 Wouter R. Ryckaert,1 Michael R. Pointer,2 Geert Deconinck,1
3 Jan Koenderink,3 and Peter Hanselaer1
1
4 Light & Lighting Laboratory, ESAT/ELECTA, KU Leuven, Heverlee, Belgium
2
5 Colour, Imaging and Design Research Centre, University of Leeds, Leeds, UK
3
6 Experimental Psychology, KU Leuven, Leuven, Belgium
7 *Corresponding author: martijn.withouck@kahosl.be

8 Received December 19, 2012; revised April 16, 2013; accepted April 30, 2013;
9 posted May 1, 2013 (Doc. ID 181968); published 0 MONTH 0000
10 1 The perception of brightness of unrelated self-luminous colored stimuli of the same luminance has been inves-
11 tigated. The Helmholtz–Kohlrausch (H-K) effect, i.e., an increase in brightness perception due to an increase in
12 saturation, is clearly observed. This brightness perception is compared with the calculated brightness according to
13 six existing vision models, color appearance models, and models based on the concept of equivalent luminance.
14 Although these models included the H-K effect and half of them were developed to work with unrelated colors,
15 none of the models seemed to be able to fully predict the perceived brightness. A tentative solution to increase the
16 prediction accuracy of the color appearance model CAM97u, developed by Hunt, is presented. © 2013 Optical
Society of America
17 OCIS codes: (330.1690) Color; (330.5020) Perception psychology; (330.5510) Psychophysics.
18 http://dx.doi.org/10.1364/JOSAA.99.099999

19 1. INTRODUCTION luminance (LEq;CIE [4] and LEq;Nay [3]) described in the next 53
20 Brightness is an attribute of visual perception according to section, include the H-K effect. Another category of stimuli 54
21 which an area appears to emit, or reflect, more or less light is unrelated colors, which are colors perceived to belong to 55
22 [1]. Luminance, defined as the luminous flux per unit pro- areas seen in isolation from any other colors [14,15]. Typical 56
23 jected area, transmitted by an elementary beam passing examples are self-luminous stimuli, light sources, railway, avi- 57
24 through a given point and in a given direction [1], can be con- ation and marine signal lights, traffic lights, and street lights 58
25 sidered as the photometric quantity most closely related to viewed with a dark surround, e.g., on a dark night. With the 59
26 brightness. However colored stimuli of equal luminance do implementation of light-emitting-diodes (LEDs) in signaliza- 60
27 not necessarily appear equally bright. The complex relation- tion and in architectural lighting applications, saturation lev- 61
28 ship between luminance and brightness—conceptualized in els are reaching much higher values. The predictive power of 62
29 a brightness-to-luminance ratio (B/L)—has been extensively the few existing models (CAM97u [16], ATD01 [17] and 63
30 studied [2–5]. Deviations from unity of the B/L ratio have been CAMFu [18]), detailed in the next section, to describe the 64
31 observed in heterochromatic brightness matches of colored brightness for this category of stimuli including the H-K effect 65
32 stimuli. They can be caused by using the wrong V λ-function has not yet been investigated systematically. 66
33 in the present standard system of photometry or a failure of In this study, nine observers have evaluated the brightness 67
Abney’s proportionality and additivity laws [6]. Experimental of 58 unrelated self-luminous colored stimuli with a luminance 68
34
of 51 cd∕m2 and surrounded by a dark background. The spec- 69
35 evidence has shown the latter to be the case in direct hetero-
tral radiance of the stimuli has been measured, and observer 70
36 chromatic brightness matching [7–11]. Although the relation-
data of the brightness have been collected. The correlation 71
37 ship between luminance and brightness can be described to a
between the brightness perception of these observers and 72
38 first-order approximation by a power law [12], it has become
the brightness calculated according to the models based on 73
39 clear that several other parameters, such as the luminance of
the equivalent luminance, the CAM97u model, the ATD01 74
40 the background and the colorfulness of the stimulus are in-
model, and the CAMFu model has been investigated. Finally, 75
41 volved as well. The effect of colorfulness or saturation on per-
a tentative improvement of the brightness prediction of 76
42 ceived brightness is referred to as the Helmholtz–Kohlrausch
CAM97u is proposed. 77
43 2 (H-K) effect [6]. It states that highly saturated colors appear
44 brighter than those of low saturation, even when they are
45 equal in luminance [3,13]. 2. VISION MODELS FOR PREDICTING 78
46 In the past, a number of models have been developed to BRIGHTNESS PERCEPTION 79
47 describe the brightness of a stimulus. Generally most models The vision models mentioned in the introduction are exten- 80
48 deal with related colors, which are perceived to belong to sively described below. 81
49 areas seen in relation to other colors [14]. Typical examples
50 are colors produced by matte objects reflecting light (object A. CAM97u 82
51 or surface colors). However, only two of the models for CAM97u is a color appearance model (abbreviated as CAM 83
52 related colors, those based on the concept of equivalent hereafter) for unrelated colors developed by Hunt [19]. CAMs 84

1084-7529/13/060001-01$15.00/0 © 2013 Optical Society of America


2 J. Opt. Soc. Am. A / Vol. 30, No. 6 / June 2013 Withouck et al.

85 try to link the experimentally measurable optical properties of displayed on a CRT [18]. Two stimulus sizes (10° and 0.5°) and 137
86 stimuli and their corresponding perceptual attributes, such as stimuli with luminance values between 0.1 and 60 cd∕m2 were 138
87 brightness, hue, colorfulness, lightness, chroma, and satura- used. For brightness evaluation a magnitude scaling method 139
88 tion under varying conditions by taking into account some [23] was adopted. Colors appeared more colorful (Hunt effect 140
89 of the physiological processes that occur in the human visual [15]) and brighter at higher luminance levels than at lower lu- 141
90 system [20]. In general, CAMs, including the most recent minance levels and increasing the stimulus size generally in- 142
91 CIECAM02 [21], are designed for object colors (related col- creased both brightness and colorfulness. The data were 143
92 ors) and do not implement the H-K effect. Most of the models compared with the predictions of CAM97u and CIECAM02, 144
93 use the tristimulus values and luminance of the stimulus, the even though the latter model is in principle not applicable 145
94 reference white, the background, and surround as input to unrelated colors. An alternative CAM for unrelated colors 146
95 parameters. under photopic and mesopic conditions, in this paper indi- 147
96 Only CAM97u, which can be considered as an extension to cated as CAMFu and highly inspired by CIECAM02, was de- 148
97 Hunt’s model for related colors [16], was developed particu- veloped. In the CAMFu model, the brightness correlate QCAMFu 149
98 larly for unrelated colors and implements the H-K effect. The is given by 150
99 input parameters of CAM97u are the XY Z tristimulus values
100 of the stimulus and the conditioning field, which is the stimu- M CAMFu
QCAMFu  ACAMFu  ; (3)
101 lus that is seen just before the test stimulus. In addition, the 100
102 photopic and scotopic luminance of the stimulus, the adapting
103 field, and the conditioning field are required. While the pho- with ACAMFu and M CAMFu the achromatic contribution and 151
104 topic luminance is used throughout the whole model (lumi- colorfulness as calculated from the CAMFu model. 152
105 nance adaptation, chromatic adaptation, photopic part of
106 the achromatic signal…) the scotopic luminance is only used D. Equivalent Luminance Nayatani (Leq;Nay ) 153
107 to calculate the scotopic part of the achromatic signal. In the Brightness has also been modeled based on the concept of 154
108 model, the brightness is predicted by using the equation: equivalent luminance, which is defined as the photopic lumi- 155
nance of a previously determined common reference stimulus 156
   that matches the test stimulus (object or self-luminous) in
M CAM97u 0.9 157
QCAM97u  1.1⋅ ACAM97u  ; (1) terms of brightness [4]. Two such models have been devel- 158
100
oped, one by Nayatani (Leq;Nay ) and one by the CIE (Leq;CIE ). 159
109 where QCAM97u is the brightness, ACAM97u the achromatic sig- The effects of the surround, background, and field of view are 160
110 nal, and M CAM97u the colorfulness [19]. The inclusion of ignored; the H-K effect is however taken into account. Both 161
111 M CAM97u in the expression for brightness represents the H-K models apply in principle only to related colors. 162
112 effect. Nayatani proposed two methods that take into account the 163
H-K effect for calculating the equivalent luminance [3]: the 164
113 B. ATD01 variable-achromatic-color (VAC) and the variable-chromatic- 165
114 The ATD01 model is a color vision model, developed by Guth color (VCC) method, given by Eqs. (4) and (5), respectively: 166
115 3 [17], built on the theoretical ideas of Helmholtz, Hering, von
Kries, and Mueller [22]. It is developed to predict the bright- Leq;Nay VAC
116 γ VAC   0.4462⋅1  f−0.1340qθ
117 ness, saturation, and hue of unrelated colors and to predict a L
118 wide range of vision science data on phenomena, such as  0.0872K Br La g⋅suv x; y  0.30863 (4)
119 chromatic discrimination, absolute thresholds, the Bezold–
120 Brücke hue shift, the Abney effect, heterochromatic bright-
121 ness matching, light adaptation, and chromatic adaptation Leq;Nay VCC
γ VCC   0.4462⋅1  f−0.8660qθ
122 [14]. In the ATD model, the XY Z tristimulus values are trans- L
123 formed into LMS cone responses. These LMS signals are gain-  0.0872K Br La g⋅suv x; y  0.30863 : (5)
124 controlled and undergo a second transformation to yield an
125 achromatic (A) and two chromatic or opponent signals The presence of the saturation suv x; y of the test stimulus 167
126 (red–green or T, blue–green or D). These A, T, and D signals indicates inclusion of the H-K effect. The function qθ de- 168
127 go through a compressive nonlinearity and are finally used to scribes the impact of the hue angle θ on the H-K effect. 169
128 calculate the perceptual attributes brightness, hue, and satu- K Br La  accounts for the increase of the H-K effect when 170
129 ration. The brightness QATD is calculated as quadrature sum of the adapting luminance of chromatic object colors is raised. 171
130 the A, T, and D signals: In the VAC method, the luminance of the reference achro- 172
matic color is changed in order to match the colored stimuli. 173
QATD  A2  T 2  D2 0.5 : (2) In the VCC method, the luminance of the colored stimuli is 174
changed in order to match the achromatic reference. 175
131 As the achromatic A value approximately corresponds to the
132 luminance Y , this model deals with the H-K effect by adding E. Equivalent Luminance CIE (Leq;CIE ) 176
133 the chromatic T and D values to the brightness prediction [8]. This model calculates a brightness-related equivalent lumi- 177
nance by using four parameters: the photopic luminance L, 178
134 C. Fu (CAMFu) the scotopic luminance L0 , an achromatic adaptation coeffi- 179
135 Fu investigated the color appearance of unrelated colors cient a, and the chromatic contribution c. The achromatic 180
136 under photopic and mesopic levels of adaptation using stimuli adaptation contribution a takes into account the so-called 181
Withouck et al. Vol. 30, No. 6 / June 2013 / J. Opt. Soc. Am. A 3

182 Purkinje effect, which causes a shift in the sensitivity of the found to gradually decrease (to approximately 20% of the 223
183 human eye toward the blue end of the visible spectrum at low average) from the center to the edge. As the human eye is in- 224
184 luminance levels, while the chromatic contribution c allows sensitive to low spatial frequencies [28], observers were not 225
185 for the H-K effect. This chromatic contribution changes with aware of this variation. The background, consisting of a black 226
186 the luminance level and the chromaticity coordinates of the curtain, provided a 0.01 cd∕m2 adaptive field extending to 35° 227
187 stimulus. A formula for the general equivalent luminance around the 10° stimuli. 228
188 Leq;CIE for related colors has been proposed [4]:
4. EVALUATION OF BRIGHTNESS 229
Leq;CIE  La :L0 1−a :10c : (6)
In the psychophysical experiment, observers were asked to 230
189
evaluate the perceived brightness of the stimuli using a mag- 231
190 This equation has been based on visual data gathered from
nitude estimation method. With this method, a numerical and 232
191 several studies [8,13,24–26] and Eq. (6) has been tested by
scalable result for the perceptual attribute under test, in this 233
192 matching experiments [5]. The supplementary system of pho-
case brightness, can be obtained directly [21,29–31]. The ex- 234
193 tometry as described by Eq. (6) was originally developed
periment started by showing a reference achromatic stimulus 235
194 based on the 2° quantities, except for the scotopic luminance,
with chromaticity close to that of illuminant D65 (u010 , 236
195 but can also be used for a centrally fixed 10° field.
v010  0.1979, 0.4695) and a luminance equal to that of the col- 237
ored test stimuli, i.e., 51 cd∕m2 . The color difference ΔEu0 v0 238
196 3. EXPERIMENTAL SETUP between the reference stimulus and the illuminant D65 was 239
197 In this study, the brightness of unrelated colors was investi- 0.00234. To keep the colored stimuli unrelated, the presenta- 240
198 gated using a specially designed viewing room [see Fig. 1(a)]. tion of a reference stimulus shown in temporal juxtaposition 241
199 The dimensions of the room are 3 m wide by 5 m long by 3.5 m with the test stimulus, possibly inducing small memory errors, 242
200 high. The walls, ceiling, and floor were covered with black was preferred to a reference stimulus presented at the same 243
201 curtains, gray panels, and a grayish black carpet, respectively. time in adjacent spatial locations. To this reference a fixed 244
202 At one wall, a circular tube (diameter 37 cm), containing an brightness value of 50 was attributed. After 5 s, the colored 245
203 RGB LED module and covered by a circular diffuser, was test stimulus was presented for 15 s. Just after switching to 246
204 mounted inside the wall [see Fig. 1(b)]. This circular diffuser the reference again, the observers were asked to rate the 247
205 provided a stimulus of approximately 10° FOV to the observ- brightness of the stimulus relative to the reference achromatic 248
206 ers who were seated on a fixed chair at a distance of 211 cm. stimulus. Pilot experiments had shown that it is easier to rate 249
207 The color of the stimulus was changed by controlling the in- the brightness immediately after the colored stimulus has dis- 250
208 tensity of the RGB LEDs using a DMX digital communication appeared. Furthermore, by showing the reference stimulus 251
209 network. For the experiments, 58 colored stimuli with a con- after each stimulus presentation, any errors induced due to 252
210 stant luminance of 51 cd∕m2 (standard deviation 0.80 cd∕m2 ) memory effects were minimized. Total darkness never 253
211 and a wide chromaticity gamut were carefully selected. This occurred in order to reduce the potential for temporary 254
212 luminance level was chosen in order to ensure photopic view- blindness and afterimages. Before the experiment, the 255
213 ing conditions without any glare effect. Note that the lumi-
214 nance was calculated using the standard spectral luminous
215 efficiency function V 10 λ for the CIE 10° observer [27].
216 The CIE 1976 u010 , v010 chromaticity coordinates of the stimuli,
217 as determined from spectral measurement using a spectrora-
218 diometer (MS260i Oriel instruments spectrograph) and suit-
219 able calibration, are illustrated in Fig. 2.
220 The luminance uniformity of the stimulus area was checked
221 by measurements with a two-dimensional luminance camera
222 (MURATest by Eldim). The luminance of the stimulus was

F1:1 Fig. 1. (a) Experimental room and (b) example of a stimulus under Fig. 2. Chromaticity coordinates of the 58 stimuli plotted in the CIE F2:1
F1:2 dark viewing conditions. 1976 u010 , v010 chromaticity diagram. F2:2

Table 1. Evaluation of Interobserver Agreement by Calculation of the Coefficient of Variation, CV


T1:1 Observer 1 2 3 4 5 6 7 8 9 Mean Median
T1:2 Inter-observer agreement CV (%) 8 12 7 7 7 16 8 10 21 11 8
4 J. Opt. Soc. Am. A / Vol. 30, No. 6 / June 2013 Withouck et al.

256 observers adapted to the dark viewing conditions for at least Table 2. Values of the Geometric Mean of the
257 4 5 min. The following instructions were given to each observer: Observer Brightness (Qgeom ), CIE 1976 u010 , v010
saturation (suv;10 ) and CIE 1976 u010 , v010
258 You will see 58 test stimuli. First a reference stimulus will
Hue-Angle (huv;10 ) of All the 58 Stimuli
259 be shown for 5 s. Each test stimulus is then presented for 15 s. Ordered by Hue
260 Between each of these 58 test stimuli, a reference stimulus
261 will be shown for 5 s. Give a value to the brightness of the Stimuli huv;10 suv;10 Qgeom T2:1
262 test stimulus immediately after the disappearing of this test Violet 291.82 0.53 54.65 T2:2
263 stimulus and in comparison with the reference. The refer- 289.04 0.86 59.82 T2:3
264 ence has a brightness value of 50. A value of zero represents 287.04 1.23 62.50 T2:4
265 a dark stimulus without any brightness. There is no upper 286.63 1.71 69.19 T2:5
266 limit to the value of brightness. Yellow 62.70 0.21 48.71 T2:6
267 To become familiar with the magnitude estimation method, 69.75 0.40 46.27 T2:7
268 a straightforward exercise was completed in which observers 78.04 0.65 48.81 T2:8
269 were asked to rate the length of a line in comparison with a 79.02 0.94 58.39 T2:9
270 line of length 100. A similar method is described in the ASTM 77.21 1.12 58.46 T2:10
271 International standard test method for unipolar magnitude es- Pink 17.62 0.21 50.57 T2:11
272 timation of sensory attributes [32]. In addition, a set of training 346.85 0.57 55.12 T2:12
273 stimuli, with the same hue and luminance as in the experi- 343.30 0.89 55.54 T2:13
341.18 1.26 68.07 T2:14
274 ments, was also presented to the observers, allowing them
339.18 1.83 73.26 T2:15
275 to be aware of the brightness range and to become familiar
Yellow–Green 83.22 0.37 49.30 T2:16
276 with brightness rating technique. The duration of this training
92.23 0.58 44.79 T2:17
277 session was about 45 min, while the experiment took about 97.46 0.79 50.24 T2:18
278 25 min. A small break was taken between the training and 97.34 0.95 52.34 T2:19
279 the experiment in order to reduce the influence of fatigue. 96.73 1.14 61.49 T2:20
280 Nine observers, five female and four male, with ages rang- Cyanogen–Blue 223.45 0.63 52.63 T2:21
281 ing between 23 and 30 years (average 27) participated in the 223.89 0.78 57.00 T2:22
282 experiments. All had normal color vision according to the 224.03 1.02 58.24 T2:23
283 Ishihara 24 plate Test for Color Blindness and were naïve with 224.11 1.34 62.56 T2:24
284 respect to the purpose of the experiment. To avoid irreprodu- 225.92 1.66 65.87 T2:25
285 cibility in the luminous flux and chromaticity of the LEDs in- Orange 36.05 0.46 48.19 T2:26
286 duced by thermal conditions, the 58 stimuli were presented in 31.29 0.88 51.98 T2:27
287 an identical sequence to all observers. The same sequence 29.33 1.23 46.60 T2:28
288 was also used for the optical measurements. The training ses- 29.24 1.55 52.18 T2:29
289 sion contained the same set of stimuli but in a different order. 28.24 1.98 64.98 T2:30
Teal 153.58 0.55 50.58 T2:31
155.98 0.62 50.73 T2:32
290 5. RESULTS 159.48 0.92 55.17 T2:33
161.38 1.18 60.36 T2:34
291 A. Interobserver Agreement
162.01 1.44 61.88 T2:35
292 The agreement between any two sets of data can be analyzed
Green 101.66 0.19 47.57 T2:36
293 using the coefficient of variation (CV) [Eq. (7)] [33]. For a per- 139.55 0.75 47.70 T2:37
294 fect agreement between two sets of data, the CV should be 143.33 1.14 60.10 T2:38
295 equal to zero. Interobserver agreement was assessed by cal- 145.02 1.54 68.18 T2:39
296 culating the CV values between each individual observer’s re- 139.03 1.97 76.04 T2:40
297 sults and the geometric mean of all the observers: Red 44.00 0.16 49.72 T2:41
s 15.68 0.78 50.88 T2:42
1X n
Qgeom;i − f Qobs;i 2 13.12 1.30 58.11 T2:43
CV  100 11.99 1.99 66.35 T2:44
n i1 Q̄2geom 11.40 2.74 64.15 T2:45
Pn
Qgeom;i Qobs;i 12.06 3.61 79.12 T2:46
with f  i1
Pn 2
: (7)
i1 Qobs;i
Blue 264.89 0.45 51.64 T2:47
258.16 1.43 61.18 T2:48
257.34 2.05 61.69 T2:49
298 Qobs;i represents the individual observer brightness of stimu-
256.10 2.69 71.59 T2:50
299 lus i, Qgeom;i the geometric mean of Qobs;i for all the observers, 252.28 3.04 76.95 T2:51
300 Q̄geom the arithmetic mean of Qgeom;i for all stimuli, n the num-
Cyanogen 195.84 0.29 51.56 T2:52
301 ber of evaluated stimuli, and f the factor adjusting the Qgeom;i
201.09 0.60 50.85 T2:53
302 and Qobs;i values to the same scale. 202.26 0.84 60.38 T2:54
303 The values for the interobserver agreement in this study 203.30 1.30 66.12 T2:55
304 ranged from 7% to 21% with an average of 11% and a median White 49.52 0.13 48.28 T2:56
305 of 8% (see Table 1). This result is similar to the value of 13% 49.45 0.54 47.58 T2:57
306 reported by Luo and Hunt [23] for the lightness of related col- 97.94 0.00 50.00 T2:58
307 ors and better than the value of 29% reported by Fu and 41.79 1.02 48.49 T2:59
Withouck et al. Vol. 30, No. 6 / June 2013 / J. Opt. Soc. Am. A 5

308 co-workers [18] when scaling the brightness of unrelated C. Model Performance 338
309 colors in conditions similar to those used in this study. The ability to predict the observer data for brightness has 339
been investigated for each of the previously described models. 340
310 B. Brightness Perception Before comparing their performances, some clarifications 341
311 The values of the geometric mean of the observers’ brightness about the input values required by the models are discussed. 342
312 Qgeom;i (further referred to as Qgeom ) for each of the 58 test Because of the 10° field-of-view (FOV) of the test stimuli, the 343
313 stimuli of equal luminance (51 cd∕m2 ) are given in Table 2, 10° photometric quantities have been calculated, although not 344
314 together with the CIE 1976 u10 , v10 hue-angle (huv;10 ) and all models were developed for this FOV. 345
315 the saturation (suv;10 ) of each stimuli, calculated using Eqs. (8) For CAM97u, the equi-energy stimulus SE was used for the 346
316 and (9) [27]: conditioning field and the adapting field. The luminance of the 347
conditioning field Lc and the adapting field La was taken as 348
  L2∕3 ∕200 [19]. The scotopic luminance of the stimuli Ls was
v010 − v0n;10 349
huv;10  arctan (8) calculated using the spectral luminous efficiency function 350
u010 − u0n;10
for scotopic vision V 0 λ and the scotopic luminance of the 351
conditioning field Lc;s and the adapting field La;s was taken 352
1 as 2.26Ls ∕2.262∕3 ∕200 [19]. 353
suv;10  13u010 − u0n;10 2  v010 − v0n;10 2 2 ; (9) For CAMFu, the experimental reference stimulus was taken 354
as the reference white. The tristimulus values of the stimuli X, 355
317 where u010 , v010 and u0n;10 , v0n;10 are the CIE 1976 chromaticity Y , Z and of the reference white X w , Y w , Z w were normalized 356
318 coordinates for the CIE 10° observer of the colored stimulus such that Y  Y w  100. The luminance of the reference 357
319 and the reference achromatic stimulus, respectively. Eleven white Lw and the adapting field La were respectively, set to 358
320 series with more or less constant hue can be identified. Lref and 1∕5Lref . The scotopic luminance of the stimuli Ls 359
321 Saturation values range from 0 to 3.61 (red stimulus). was calculated using the spectral luminous efficiency function 360
322 It is clear that for each hue series, there is generally an in- for scotopic vision V 0 λ. 361
323 crease in perceived brightness with saturation. A plot of Qgeom For each of the six models described in the second section, 362
324 for each stimulus versus the saturation suv;10 is given in Fig. 3. the averaged visual brightness as assessed by the observers, 363
325 As each stimulus has the same luminance, Fig. 3 clearly Qgeom , has been plotted against the calculated brightness, Q, 364
326 illustrates the H-K effect. The figure suggests a possible differ- on Fig. 4. The blue, green, red, and yellow stimuli have been 365
327 ence in the size of the H-K effect for different colors (larger for highlighted. To assess the amount of variation in brightness 366
328 red and blue, lower for yellow). Therefore the dependency of perception that is explained by each model, the coefficient 367
329 the perceived brightness on the 11 hue series has been of determination (R2 ) of the regression between the observed 368
330 investigated using a one-way ANOVA. With Qgeom as depen- and calculated brightness has been calculated. A R2 close to 1 369
331 dent variable and the 11 hue series as factor, the analysis suggests a good prediction by the model [34]. Although a lin- 370
332 showed that the observed brightness is not significantly differ- ear relation between the observed and the calculated bright- 371
333 ent for the eleven hues, F10; 43  1.347, p  0.237. Even ness is expected, the Spearman correlation coefficient [34] 372
334 though prior studies of the H-K effect showed a tendency has also been calculated. The Spearman correlation coeffi- 373
335 to be larger for reds and blues than for yellow [6], the current cient is a rank order metric (not sensitive to the potential non- 374
336 experimental data was insufficient to support this with linearity of the relation between observed and calculated 375
337 statistical significance. values of Q), having a value between −1 (perfect negative 376

F3:1 Fig. 3. Average observer brightness (Qgeom ) with SE error bars, calculated for each individual stimulus from all observer answers, plotted against
F3:2
F3:3 the CIE 1976 u010 , v010 saturation (suv;10 ) of the stimuli with the yellow, green, blue, and red stimuli highlighted. 4/C
6 J. Opt. Soc. Am. A / Vol. 30, No. 6 / June 2013 Withouck et al.

F4:1 Fig. 4. Average observer brightness (Qgeom ) with SE error bars plotted against the brightness predictions of (a) CAM97u, (b) ATD01, and
F4:2 (c) CAMFu and the predictions of the equivalent luminance of Nayatani [VAC (d) and VCC (e)], and (f) CIE. The blue, green, red, and yellow
F4:3 stimuli are highlighted.

377 correlation) and 1 (perfect positive correlation). Table 3 the low values of the Spearman correlation coefficient, and 403
378 summarizes the statistical results for each model. the low coefficient of determination (Table 3) indicate that 404
379 From Table 3, it is clear that none of the described models they are unable to predict the experimental brightness data 405
380 perform well. Although the two best models, LEq;CIE R2  (with a luminance of 51 cd∕m2 ). The brightness prediction 406
381 0.730 and LEq;Nay VACR2  0.718, included the H-K effect of CAM97u [Eq. (1)] and CAMFu [Eq. (3)] both consist of a 407
382 explicitly, they were not developed for unrelated colors, summation of an achromatic signal, which is nearly constant 408
383 which could explain their lack of predictive strength for (all stimuli have equal luminance), and a contribution of the 409
384 the constant luminance stimuli. Remarkably the VAC model colorfulness factor M, which takes into account the H-K ef- 410
385 of Nayatani’s equivalent luminance, where the achromatic fect. The failure of these two models to predict the perceived 411
386 color is changed to match the colored stimuli, performs better brightness for the self-luminous stimuli in this study might be 412
387 than the VCC model (R2  0.598), although it should be less attributed to a poor implementation of this colorfulness 413
388 applicable to the method used in this experiment, with a
389 constant achromatic stimulus.
390 While the one-way ANOVA showed that there was not a sig- Table 3. Overview of the Correlation between the
391 nificant difference between Qgeom for the 11 hue series, the Average Brightness Data of the Observers and the
392 same analysis with LEq;CIE , LEq;Nay VAC and LEq;Nay VCC Predictions of the Vision Models Together with
393 as dependent variables (Table 3) indicated that there is a dif- the One-Way ANOVA of the 11 Hue Series
394 ference between the brightness predicted by these models for One-Way ANOVA T3:1
395 the 11 hues. In fact, the models predicted a difference between
Model R2 Spearman r F3; 17 p T3:2
396 the brightness for these hues while this was not observed
397 by the observers. QCAM97u 0.131 −0.178 10.128 0.000 T3:3
398 The one-way ANOVA with the brightness prediction of the QATD01 0.338 0.462 6.167 0.000 T3:4
399 other models—CAM97u, ATD01, and CAMFu—also showed a QCAMFu 0.061 0.357 13.487 0.000 T3:5
400 difference between the predicted brightness for the 11 hue LEq;Nay (VAC) 0.718 0.881 2.390 0.024 T3:6
401 series. Although these have been developed for unrelated LEq;Nay (VCC) 0.598 0.769 5.510 0.000 T3:7
LEq;CIE 0.730 0.906 2.319 0.028 T3:8
402 colors and include the H-K effect, the results from the ANOVA,
Withouck et al. Vol. 30, No. 6 / June 2013 / J. Opt. Soc. Am. A 7

increased colorfulness contribution should be determined, ad- 451


ditional visual data at different luminance levels are required 452
in order to propose a better model. 453

6. CONCLUSIONS 454
In a psychophysical experiment the brightness perception of 455
unrelated self-luminous colored stimuli with a constant lumi- 456
nance was investigated. The ability of six vision models to pre- 457
dict the observed brightness was evaluated using the 458
coefficient of determination, the Spearman correlation coeffi- 459
cient, and an ANOVA analysis. Although the models included 460
the H-K effect and half of them were developed to work with 461
unrelated colors, none of the models seemed to be fully able to 462
predict the perceived brightness. The expected linear relation- 463
ship between the observed and predicted brightness was not 464
achieved. 465
However, the brightness prediction of CAM97u and CAMFu 466
both consist of a summation of an achromatic signal, which is 467
nearly constant (all stimuli have equal luminance), and a con- 468
tribution of the colorfulness factor M, to take the H-K effect 469
F5:1 Fig. 5. Average observer brightness (Qgeom ) with SE error bars plot- into account. The failure of these two models to predict the 470
F5:2 ted against the colorfulness predictions of CAM97u. The blue, green, perceived brightness for the self-luminous stimuli in this study 471
F5:3 red, and yellow stimuli are highlighted.
might thus be attributed to a poor implementation or an 472
underestimation of this colorfulness factor. The former is 473
414 factor. In fact, the correlation between M and Qgeom can be most likely the case for the Fu model as the correlation be- 474
415 used to investigate the implementation of the H-K effect. tween the observed brightness and its predicted colorfulness 475
416 The correlation between M CAMFu and Qgeom R2  is rather low and the model is based on CIECAM02, which is 476
417 0.483 and Spearman r  0.607 indicates that the H-K effect unable to handle unrelated colors. However, CAM97u, specifi- 477
418 is poorly predicted. This failure of CAMFu might be attributed cally designed for unrelated colors, showed a good correlation 478
419 to the use of a colorfulness factor M CAMFu based on the between its colorfulness factor and the observed brightness 479
420 CIECAM02 model [21], which was not developed to handle suggesting that an increase of the contribution of the colorful- 480
421 unrelated colors. ness factor in the calculation of the brightness might lead to a 481
422 The correlation between M CAM97u and Qgeom R2  better model. In order to determine the exact magnitude and 482
423 0.770 and Spearman r  0.859 shows a better prediction of possible luminance dependence of this increased colorfulness 483
424 the H-K effect (Fig. 5). A more detailed analysis of the hue contribution, further research at different luminance levels 484
425 dependency of M CAM97u using a one-way ANOVA analysis in- will be performed. 485
426 dicates that there is no significant difference between the pre-
427 dicted brightness for the 11 hues, F10; 43  1.071, ACKNOWLEDGMENTS 486
428 p  0.405, which is in agreement with the observers not rating The authors would like to thank the Research Council 487
429 the hue series differently. Also a customized ANCOVA was of the KU Leuven for supporting this research project 488
430 performed to test the assumptions of homogeneity of regres- (STIM-OT/11/056). 489
431 sion slopes between the observed brightness and the colorful-
432 ness of CAM97u for the 11 hue series [34] by including, next to
433 the main effect of Qgeom and M CAM97u , the interaction term. REFERENCES 490
434 This customized analysis with Qgeom as dependent variable, 1. CIE, International Lighting Vocabulary (CIE Central Bureau, 491
2011). 492
435 the 11 hues as fixed factors and M CAM97u as covariate, showed 2. C. L. Sanders and G. Wyszecki, “Correlate for brightness in 493
436 that the effect of the interaction term between Qgeom and terms of CIE color matching data,” in CIE Proceedings 494
437 M CAM97u is not significant (F10; 43  1; 114, p  0.382). In 15th Session (CIE Central Bureau, 1963). 495
438 fact the assumption of homogeneity of the slopes between 3. Y. Nayatani, “Simple estimation methods for the Helmholtz— 496
Kohlrausch effect,” Color Res. Appl. 22, 385–401 (1997). 497
439 Qgeom and M CAM97u for each hue is confirmed. The relationship 4. CIE, Supplementary System of Photometry (CIE Central 498
440 between the observed brightness and the colorfulness of Bureau, 2011). 499
441 CAM97u is thus similar for the eleven hues. 5. CIE, Testing of Supplementary Systems of Photometry (CIE 500
442 The H-K effect can be thus predicted by M CAM97u while the Central Bureau, 2001). 501
443 achromatic factor ACAM97u is nearly constant for all 58 stimuli 6. G. Wyszecki and W. S. Stiles, Color Science, 2nd ed. (Wiley, 502
1982), p. 410. 503
444 and QCAM97u does not correlate with the observed brightness. 7. S. L. Guth, N. J. Donley, and R. T. Marrocco, “On luminance 504
445 This suggests that the performance of CAM97u could be con- additivity and related topics,” Vis. Res. 9, 537–IN531 (1969). 505
446 siderably improved just by increasing the contribution of this 8. S. L. Guth and H. R. Lodge, “Heterochromatic additivity, foveal 506
447 colorfulness. In fact, increasing the contribution of the color- spectral sensitivity, and a new color model,” J. Opt. Soc. Am. 63, 507
450–462 (1973). 508
448 fulness in Eq. (1) will also increase the correlation between 9. R. M. Boynton and P. K. Kaiser, “Vision: the additivity law made 509
449 Qgeom and QCAM97u . However, because ACAM97u and M CAM97u to work for heterochromatic photometry with bipartite fields,” 510
450 are not totally independent and the magnitude of the Science 161, 366–368 (1968). 511
8 J. Opt. Soc. Am. A / Vol. 30, No. 6 / June 2013 Withouck et al.

512 10. G. Wagner and R. M. Boynton, “Comparison of four methods of 22. P. Capilla, M. J. Luque, J. Gómez, and A. Palomares, “On satu- 545
513 heterochromatic photometry,” J. Opt. Soc. Am. 62, 1508–1515 ration and related parameters following Guth’s ATD colour- 546
514 (1972). vision model,” Color Res. Appl. 26, 305–321 (2001). 547
515 11. P. K. Kaiser and G. Wyszecki, “Additivity failures in heterochro- 23. M. R. Luo, A. A. Clarke, P. A. Rhodes, A. Schappo, S. A. R. 548
516 matic brightness matching,” Color Res. Appl. 3, 177–182 (1978). Scrivener, and C. J. Tait, “Quantifying colour appearance. Part 549
517 12. S. S. Stevens, “On the psychophysical law,” Psych. Rev. 64, I. Lutchi colour appearance data,” Color Res. Appl. 16, 166–180 550
518 153–181 (1957). (1991). 551
519 13. H. Yaguchi and M. Ikeda, “Subadditivity and superadditivity in 24. K. Sagawa and K. Takeichi, “System of mesopic photometry for 552
520 heterochromatic brightness matching,” Vis. Res. 23, 1711–1718 evaluating lights in terms of comparative brightness relation- 553
521 (1983). ships,” J. Opt. Soc. Am. A 9, 1240–1246 (1992). 554
522 14. M. D. Fairchild, Color Appearance Models, 2nd ed., Wiley-IS&T 25. Y. Nakano, “A model for brightness perception and its 555
523 Series in Imaging Science and Technology (Wiley, 2005). application to individual data,” Jpn. J. Opt. 21, 705–716 556
524 15. R. W. G. Hunt and M. R. Pointer, Measuring Colour, 4th ed., (1992). 557
525 Wiley-IS&T Series in Imaging Science and Technology (Wiley, 26. D. A. Palmer, “Standard observer for large-field photometry at 558
526 2011). any level,” J. Opt. Soc. Am. 58, 1296–1298 (1968). 559
527 16. R. W. G. Hunt, “Revised colour-appearance model for 27. CIE, Colorimetry (CIE Central Bureau, 2004). 560
528 related and unrelated colours,” Color Res. Appl. 16, 146–165 28. T. N. Cornsweet, Visual Perception (Academic, 1970). 561
529 (1991). 29. F. B. Leloup, M. R. Pointer, P. Dutré, and P. Hanselaer, 562
530 17. S. L. Guth, “ATD01 model for color appearances, color “Luminance-based specular gloss characterization,” J. Opt. 563
531 differences and chromatic adaptation,” in 9th Congress of the Soc. Am. A 28, 1322–1330 (2011). 564
532 International Colour Association, Proceedings of SPIE Vol. 30. G. A. Gescheider, “Psychophysical scaling,” Annual Rev. Psych. 565
533 4421 (SPIE, 2002). 39, 169–200 (1988). 566
534 18. C. Fu, C. Li, G. Cui, M. R. Luo, R. W. G. Hunt, and M. R. Pointer, 31. W. S. Torgerson, Theory and Methods of Scaling (Wiley, 567
535 “An investigation of colour appearance for unrelated colours 1958). 568
536 under photopic and mesopic vision,” Color Res. Appl. 37, 32. A. International, Standard Test Method for Unipolar Magnitude 569
537 238–254 (2012). Estimation of Sensory Attributes (West Conshohocken, 570
538 19. R. W. G. Hunt, Measuring Colour, 3rd ed. (Fountain, 1998), 2012). 571
539 pp. 239–246. 33. P. A. García, R. Huertas, M. Melgosa, and G. Cui, “Measurement 572
540 20. M. R. Luo and R. W. G. Hunt, “The structure of the CIE 1997 of the relationship between perceived and computed 573
541 Colour Appearance Model (CIECAM97s),” Color Res. Appl. color differences,” J. Opt. Soc. Am. A 24, 1823–1829 574
542 23, 138–146 (1998). (2007). 575
543 21. CIE, A Colour Appearance Model for Colour Management Sys- 34. A. Field, Discovering Statistics Using SPSS, 3rd ed. (SAGE, 576
544 tems: CIECAM02 (CIE Central Bureau, 2004). 2009). 577
Withouck et al. Vol. 30, No. 6 / June 2013 / J. Opt. Soc. Am. A 9

Queries

1. AU: Please provide the postal code for all the affiliations.

2. AU: Please check the edits made in the sentence “The effect of colorfulness...”

3. AU: Please check the author names which does not match with the Ref. 22 in the list.

4. AU: Please check para style.

You might also like