Effects of Peripheral Gradient of Color Saturation on the Feeling of Being Dazzled

Introduction

A uniform white region surrounded by an area with luminance gradient, as shown in the images in Figure 1, appears to emit light even when printed on paper and not actually emitting light. Such self-luminous appearance and brightness enhancement that occurs when a region is surrounded by a luminance gradient is called the glare effect (Zavagno & Caputo, 2005; Lu, Zavagno, & Liu, 2006). The glare effect does not depend on the details of the stimulus configuration, and a number of variants that appear glowing have been reported (e.g., the sun figure composed of streaks made of black dots (Kennedy, 1976), glowing diamond shape (Bressan, 2001), and luminous cross shape and glowing Kanizsa triangle (Zavagno, 1999). A central uniform region and surrounding luminance gradations in which luminance decreases from the center to the periphery seem to suffice to create a luminous appearance. The shape of the central region and the surrounding region has little effect on the glowing appearance, and the shape of the surrounding region and the luminance profile of the luminance gradient do not noticeably affect the glare effect (Zavagno, 1999). Such stimuli of a uniform central region surrounded by regions with luminance gradient that have glowing appearance do not have a common name. In this work, we will refer to these as “glare images.” Luminosity and brightness perception evoked by glare images have been investigated in a number of studies (e.g., Leonards, Troscianko, Lazeyras, & Ibanez, 2005; Tamura, Nakauchi, & Koida, 2016; Zavagno & Caputo, 2001, 2005). However, how the color of glare images affects brightness and luminosity perception is not yet known. Although glare effects are known to be robust to variation of the stimulus configuration based on simple observations, the effects of stimulus configuration on perceived luminosity have yet to be systematically examined.

OPEN IN VIEWER

Figure 1.

The upper row shows the three types of color saturation gradients used in the experiment. The lower row shows the five colors used for the annulus. Note: Please refer to the online version of the article to view the figures in colour.

Glare images usually appear in photographs of light sources or specular reflections of light on glossy or metallic surfaces, as illustrated in Figure 2. The central uniform white is clipped or blown-out white, as would be caused by overexposure or a saturated response in the camera. Such saturated responses also occur in the human visual system when we view a bright light source directly. The peripheral gradation of luminance occurs because of light scattering in the eye or a camera lens and for other optical reasons (Vos, 1984). The resemblance of the glare images to photographic images of light sources, such as those in Figure 2, is considered one of the reasons for the perception of glow. Such glare images often appear in photographs and movies and have also been used in computer graphics (Kakimoto, Matsuoka, Nishita, Naemura, & Harashima, 2005; Spencer, Shirley, Zimmerman, & Greenberg, 1995). Images mimicking clipped white with a surrounding flare with luminance gradient are called simply “glare” in the field of computer graphics. As we cannot present very high luminance on a typical monitor, we use glare images in computer graphics to depict patterns with high luminance, such as specular reflection of a light source from a glossy or metallic surface. Glare images have also been used in art. For example, the artist Christian Riese Lassen (2018) drew many fantastic pictures with glare images of the moon and sun and was known for this technique of representing brilliant light. Glare images have also been used in animation films. For example, glare images were abundantly used in the Japanese animation movie, Kimi no Na wa (Your Name in English), directed by Makoto Shinkai (Kimi no Na wa, n.d.) and praised by many animation fans all over the world. This animation film was well known for its realistic, detailed drawings, and the glare images seem to contribute to its sense of reality.

OPEN IN VIEWER

Figure 2.

Two photographs of light sources taken by the camera of a tablet computer. The central region is clipped white, which is a result of response saturation. Graduation appears around the central region, and the color saturation in the surrounding region increases from the central region to the periphery. Note: Please refer to the online version of the article to view the figures in colour.

In studies of illumination, glare indicates the sensation produced by strong light to cause discomfort, loss in visual performance, and visibility (Carlucci, Causone, De Rosa, & Pagliano, 2015). ¹ Studies of glare have predominantly focused on two themes. One is disability glare, which means the degradation of visibility by scattering of light in the eye (Vos, 2003a). In the studies of disability glare, scattering of light in the eye and invisibility due to it have been modeled (e.g., Aslam, Haider, & Murray, 2007; Holladay, 1926; Stiles & Crawford, 1937; Vos, 1984, 2003b). The other is discomfort glare, which indicates discomfort caused by bright peripheral light. Models to predict discomfort caused by illumination have been presented (e.g., Chauvel, Collins, Dogniaux, & Longmore, 1982; Einhorn, 1979; Guth, 1963; Petherbridge & Hopkinson, 1950). However, these models of disability and discomfort glare generally do not take light wavelength into account, except for the luminous efficiency function for the definition of luminance.

Vos (2003a) has pointed out another kind of glare phenomenon, dazzling glare, which is reaction to overexposure of light and may include light avoidance or pain. Dazzling may be implicitly included as a kind of discomfort, and the dazzling may be a mediator of discomfort due to strong light. However, dazzling is conceptually different from discomfort, for dazzling seems able to occur without discomfort, and vice versa. However, there are few studies of dazzling glare in the field of illumination studies.

Glare images evoke not only perception of glow but also a feeling that the eyes are dazzled. Because light input from glare images on a monitor is generally not very strong, observers’ eyes are not actually dazzled and they do not generally feel pain from the stimulation. Thus, strictly speaking, it may not be a case of dazzling glare. However, we often feel the eyes to be somewhat dazzled and seek to avert them from glare images. Here, we call such impressions caused by glare images the feeling of being dazzled (mabushisa in Japanese). Glare images have been used in many fields, such as motion pictures and computer graphics, and invoking the feeling in observers of being dazzled seems to be one of the reasons for the use of glare images. For example, glare images can give the illusory impression of entering a bright room from a dark place. The terminology used in this study and phenomena related to glare are summarized in Figure 3.

OPEN IN VIEWER

Figure 3.

Illustration of the phenomena related to glare and self-luminosity perception. Actual strong light as from the sun causes several phenomena in the observer such as disability or invisibility, discomfort, dazzling, and self-luminous perception (note that actual light is drawn as a schematic illustration of the sun). The first three are treated as glare in illumination studies. The scattering of light in the eye is considered a cause of these phenomena. The image mimicking the scattering of strong light, which we call glare image, would not cause disability and appears self-luminous. The glare image may also lead to some discomfort and evokes the feeling of being dazzled, though it does not generally cause pain that is accompanied by dazzling. Discomfort and dazzling may be mutually dependent, and the perception of self-luminosity may be a reason for the feeling of being dazzled.

Although the feeling of being dazzled caused by glare images has not been so intensively studied so far, a few studies have been conducted. Hanada (2012) examined the effects of luminance profile in the peripheral gradational region on the feeling of being dazzled. Images consisting of a central disk surrounded by an annulus with a luminance gradient, similar to the images presented in Figure 1, were used. It was found that an annulus’s luminance profile that makes the border between the disk and the annulus vague reduced the feeling of being dazzled. This finding suggests that indistinctness between the disk and the annulus weakens the feeling of being dazzled. In another study, Hanada (2015) varied the colors of the disk and the annulus to investigate the influence of colors on the feeling of being dazzled. He reported that pink and light blue annuluses evoked a stronger dazzling feeling than gray, green, and yellow annuluses. It was also found that the feeling of being dazzled tended to be weaker when the disk and the annulus had the same color than when they had different colors. This also suggests that distinctness of the central disk from a surrounding annulus enhances the feeling of being dazzled. The glare image of a disk surrounded by an annulus with a luminance gradient somewhat resembles the shading of a round object. Thus, the visual system may unconsciously process the image not only as a shaded object but also as a luminous light source. The distinctness of the border between the disk and the annulus may reduce the object-like appearance and enhance the perceived luminosity, which may lead to a stronger feeling of being dazzled (Hanada, 2012, 2015).

An image of a light source taken by a camera generally consists of a clipped white region surrounded by a region with a luminance gradient. R, G, and B sensors of a camera cannot respond infinitely, and their responses are saturated at some light intensity levels. If a light source emits strong chromatic light, R, G, and B sensors respond maximally in the same way as they respond to white light. Thus, the responses of the R, G, and B sensors indicate white. This is the process that causes clipped white in the central region. The surrounding region of the image occurs due to the scattering of light in the camera. The scattering light is weaker than the original light, and the sensor responses to the scattering light should become unsaturated, and the R, G, and B sensors can encode colors other than white. Because the scattering light weakens gradually from the central region to the periphery, the responses of R, G, and B sensors become gradually unsaturated from the center to the periphery, and the encoded color becomes closer to the original color of the light source. Thus, color saturation should increase and luminance should decrease from the central border to the periphery. Similar processes should also arise for the photoreceptors in human eye. Actual photographs of light sources, as shown in Figure 2, clearly illustrate this point; color saturation in the halo-like region increases from the center to the periphery. Note that the change of the registered or perceived color is not due to physical change of wavelength composition of light scattered in the camera or eye, but due to the response saturation of the color registering mechanism. Thus, in the case of the eye, a sensory phenomenon is involved. When achromatic light is emitted instead, the surrounding halo gradually blends into the background, which generally leads to an increase in color saturation from the center to the periphery. Thus, an increase in color saturation from the central area of the border to the periphery in the surrounding region should be natural for the image of a light source.

In a previous study of Hanada (2015), however, the whole annulus had the same chromaticity, which is somewhat unnatural in this respect. In this study, the type of color saturation gradient in the periphery was varied, and its effects on the feeling of being dazzled were examined using an achromatic central region. Three types of saturation gradient were used (Figure 1). For increasing saturation gradient, color saturation increased from the center to the periphery. This is the natural saturation gradient for glare images. For decreasing saturation gradients, color saturation decreased from the center to the periphery, so that chromaticity abruptly changed at the center-periphery border. For the no-saturation gradient, color saturation in the peripheral region was uniform, as in the study of Hanada (2015).

Thus, we have two potential factors that may strengthen the feeling of being dazzled: discontinuity between the central and peripheral regions, and naturalness of the color saturation gradient. Depending on which factor is effective for the feeling of being dazzled, here we present two incompatible hypotheses about the effects of the saturation gradients on the feeling of being dazzled. Hypothesis A: The feeling of being dazzled will be stronger when there is a chromaticity border between the center and the periphery than when there is not; the feeling of being dazzled will be stronger for cases of decreasing and no-saturation gradients than for increasing saturation gradients. Hypothesis B: The feeling of being dazzled will be stronger for a natural saturation gradient than an unnatural saturation gradient; it will be strongest for increasing saturation gradients among the three types of saturation gradients. Note that Hypothesis B is not expected to hold in the condition where the center luminance is lower than the periphery maximum luminance, because the center cannot then be interpreted as clipped white, which is unnatural for an image of a light source regardless of the saturation gradient in the periphery. This study tested the two hypotheses by conducting an experiment in which the types of color saturation gradient were systematically varied.

Previous studies on the feeling of being dazzled (Hanada, 2012, 2015) used a disk annulus shape similar to that shown in Figure 1. However, in a number of studies of luminosity perception or glare effect (e.g., Facchin, Daini, & Zavagno, 2017; Leonards et al., 2005; Zavagno & Caputo, 2001, 2005) and related studies on brightness perception (e.g., Agostini & Galmonte, 2002), a cross-like shape composed of a central uniform square surrounded by four squares with luminance gradients was used as stimulus. Hence, this study also used this shape as a stimulus configuration. A photographic image of a light source often has several streaks or angular variations of luminance in the peripheral region, as can be observed in the pictures in Figure 2. Hence, we also used a stimulus configuration with angular variation. We varied the stimulus configuration in addition to the saturation gradient to observe how different stimulus configurations affect the effects of saturation gradients on the feeling of being dazzled.

Methods

Stimuli

The following is a description of the three stimulus configurations that were used in the study (see Figure 5).

Disk annulus

The stimulus consisted of a disk surrounded by an annulus. The radius of the disk was 2.4°, and the annulus spanned from 2.4° to 7.2°. This stimulus was similar to that used in the study by Hanada (2012); however, the size of the stimulus was different. The annulus was larger than those used in Hanada (2012) because the stimulus size had been designed so that all the stimulus configurations were comparable. The disk had uniform luminance, which was set to 32.5 or 65.0 cd/m². The luminance of each pixel in the annulus linearly decreased from the inner border to the periphery as a function of the distance from the central disk to the pixel. The luminance at the outer boundary of the annulus was the same as that of the black background. The maximum luminance at the inner boundary of the annulus (i.e., maximum luminance for the annulus) was also set to 32.5 or 65.0 cd/m².

Star

The stimulus was similar to the disk-annulus, but the luminance of the annulus was angularly modulated; the luminance of a point in the annulus was determined by the following function:

L = L o - L max - L o 1 - r / r p sin 2 ⁡ 4 θ

where L o is the luminance of the corresponding point for the disk annulus condition, L max is the periphery maximum luminance, θ is the angle between the horizontal line and the direction of the respective point, and r and r p are the distances from the center to the current point and the circumference of the peripheral region (i.e., r p  = 7.2°). The function was not determined by theoretical modeling but was selected on an ad hoc basis to yield a stimulus that looks like a star, as shown in Figure 5. Because of the factor L max - L o in the second term, this function leaves the region with the maximum peripheral region unchanged, and the luminance in the peripheral region just adjacent to the disk was not modulated. It should also be noted that the luminance of a peripheral point for the star configuration was lower than or equal to the luminance of the corresponding point for the disk annulus.

Cross

The stimulus consisted of the pattern of a central square surrounded by four squares with luminance gradient, which looks like a cross shape and has been used often in studies of luminosity perception. (The background color used in these studies was usually white; however, in this study, the background color was black. Zavagno and Caputo (2001) reported that luminosity perception occurred more easily against backgrounds with lower luminance.) The size of the central and surrounding squares was 2.4° × 2.4°. The central square had uniform luminance and the luminance of the surrounding squares decreased linearly as a function of the distance from the central square. The background of the stimulus was black (immeasurable, and much less than 0.1 cd/m²) for all the stimulus configurations.

The color of the central region was fixed as achromatic gray. ² Five colors were used for the periphery: green, pink, yellow, light blue, and gray. (We use the terms “center” and “periphery” to refer to the central disk and the annulus for the stimulus configurations of disk-annulus and star, and the center square and the surrounding four squares for the cross configuration.) The International Commission on Illumination (CIE) xy coordinates for the five colors were as follows: pink: (0.34, 0.31), green: (0.27, 0.33), yellow: (0.34, 0.35), light blue: (0.23, 0.27), and gray: (0.29, 0.30). These coordinates were also used in the study by Hanada (2015). Assuming that the CIE xy coordinates and luminance of the reference white point were (0.285, 0.296) and 89.1 cd/m², the four chromatic colors (red, green, yellow, and light blue) with 65.0 cd/m² had approximately the same chroma in the CIELAB color space, which means that the saturation of the four colors was almost identical.

Three kinds of periphery color saturation gradients were used for the periphery (Figure 1): increasing, decreasing, and none. For the increasing saturation gradient, the saturation of the color increased from the center to the periphery, such that the chromaticity also continuously changed from the center to the periphery. The color of a pixel in the peripheral region was generated by mixing achromatic gray and one of the chromatic colors described above at a ratio of p:(1−p), where p indicates the ratio of the luminance of the current pixel to the maximum annulus luminance.

For the decreasing saturation gradient, the saturation of the color decreased from the center to the periphery, and the chromaticity between the center and the periphery was discontinuous. The color of a pixel in the peripheral region was generated by mixing achromatic gray and one of the chromatic colors at a ratio of (1−p):p, the inverse of the ratio for the increasing condition.

For the none condition, the chromaticity (or CIE xy color coordinates) for the peripheral region was uniform, which means that there was no saturation gradient under this condition. The stimulus in this condition for the disk annulus was similar to that used in the study of Hanada (2015), though the size of the stimulus was slightly different. Note that chromaticity of a pixel was changed while keeping the luminance constant. For the achromatic (gray) periphery, we could not make a saturation gradient and only the no-saturation gradient was used.

Thus, for each of the stimulus configurations (disk-annulus, star, and cross), there were the following 48 stimulus conditions for the chromatic periphery: four peripheral colors (pink, green, yellow, and light blue) × three peripheral saturation gradients (increasing, decreasing, and none) × two center luminances (32.5 and 65.0 cd/m²) × two peripheral maximum luminances (32.5 and 65.0 cd/m²). For the gray periphery, there were four stimulus conditions: two center luminances × two periphery maximum luminances. In addition, there were two control conditions without the periphery: two center luminances (32.5 and 65.0 cd/m²). The total number of stimulus conditions was 48 + 4 + 2 = 54 for each of the stimulus configurations (disk-annulus, star, and cross).

Results

The data for the stimuli without a periphery and for those with a gray periphery were not included in the statistical analysis, because these conditions did not conform to the experimental design of stimuli with a chromatic periphery. In addition, because this study focuses on the gradient of color saturation, but not on the peripheral color, the data were collapsed over the peripheral colors; conditions of different peripheral colors were treated as the same condition. The conclusions of the results described here were not changed whether the peripheral colors were included in the factors or not. However, it should be noted that the light-blue periphery evoked a stronger feeling of being dazzled than the pink, green, and yellow peripheries (the grand mean ratings were 6.0 for blue, 5.7 for pink, 5.7 for green, and 5.6 for yellow).

The overall means of individual means of the ratings for each stimulus condition collapsed across periphery colors for three stimulus configurations (disk-annulus, star, and cross) are shown in Figure 4. The statistical significances shown in Figure 4 are summarized in Figure 5 with illustrative images. We examined the results under the three stimulus configurations separately. The three-way repeated-measures analysis of variance showed that the three-way interactions of Center Luminance × Periphery Maximum Luminance × Periphery Saturation Gradient were significant for all three stimulus configurations, F(2, 30) = 15.4, p < .001 for the disk-annulus; F(2, 28) = 7.0, p = .004 for the star; F(2, 28)= 4.5, p = .020 for the cross. Because the main focus of this study was on the periphery saturation gradient, the simple main effect of the periphery saturation gradient for each pair of the center and periphery luminances was examined (Table 1). If it was significant, the Ryan’s multiple comparison test was conducted. The results under the three stimulus configurations are discussed separately here.

OPEN IN VIEWER

Figure 4.

Average ratings of the feeling of being dazzled for the stimulus configuration of the disk-annulus (a), star (b), and cross (c). *p < .05. **p < .01.

OPEN IN VIEWER

Figure 5.

Summary of the significant differences in the feeling of being dazzled between periphery saturation gradients shown in Figure 4. Although the periphery color in the images was pink, the significances indicate statistically significant differences in the ratings averaged across the four periphery colors (pink, green, yellow, and light blue). Note: Please refer to the online version of the article to view the figures in colour.

OPEN IN VIEWER

Table 1.

Simple Main Effects of Periphery Chromaticity Gradient for Each Combination of Center Luminances and Periphery Maximum Luminances for the Stimulus Configurations of Disk-Annulus, Star, and Cross.

Stimulus configuration	Center and periphery luminances	df	MS	F	p	Partial η2
Disk-annulus	Center: 32.5 cd/m2, Periphery: 32.5 cd/m2	2	1.01	9.06	<.001***	0.13
	Center: 32.5 cd/m2, Periphery: 65.0 cd/m2	2	0.75	6.73	.002**	0.10
	Center: 65.0 cd/m2, Periphery: 32.5 cd/m2	2	0.18	1.64	.199	0.03
	Center: 65.0 cd/m2, Periphery: 65.0 cd/m2	2	0.97	8.62	<.001***	0.13
	Error	120	0.11
Star	Center: 32.5 cd/m2, Periphery: 32.5 cd/m2	2	1.00	6.64	.002**	0.11
	Center: 32.5 cd/m2, Periphery: 65.0 cd/m2	2	1.37	9.08	<.001***	0.14
	Center: 65.0 cd/m2, Periphery: 32.5 cd/m2	2	0.41	2.75	.068	0.05
	Center: 65.0 cd/m2, Periphery: 65.0 cd/m2	2	1.06	7.05	.001**	0.11
	Error	112	0.15
Cross	Center: 32.5 cd/m2, Periphery: 32.5 cd/m2	2	0.41	2.44	.091	0.04
	Center: 32.5 cd/m2, Periphery: 65.0 cd/m2	2	0.06	0.34	.714	0.01
	Center: 65.0 cd/m2, Periphery: 32.5 cd/m2	2	0.39	2.30	.105	0.04
	Center: 65.0 cd/m2, Periphery: 65.0 cd/m2	2	1.08	6.40	.002**	0.10
	Error	112	0.17

Note. MS = mean squares.

*p < .05. **p < .01. ***p < .001.

Discussion

In this study, the feeling of being dazzled due to glare images was examined. For the stimulus configuration of the disk-annulus, the feeling of being dazzled was higher for the increasing color saturation gradient from the center to the periphery than for the decreasing and no-saturation gradients when the center and the periphery maximum luminances were the same. This suggests that the presence of a chromaticity difference between disk and annulus enhances the feeling of being dazzled. Similar but not identical results were observed for the stimulus configuration of the star. For the cross configuration, however, quite different results were obtained; when both center and periphery had high luminance, the increasing saturation gradient evoked a stronger feeling of being dazzled compared with the other saturation gradients.

The feeling of being dazzled was weaker when the disk and the annulus were chromatically continuous than when they were discontinuous, when the disk luminance was the same as the annulus maximum luminance (Figure 5). These results support Hypothesis A that the dazzling feeling would be stronger when there was a chromatic border between the center and the periphery. However, when the disk luminance was different from the annulus maximum luminance, different results were obtained. Because there was a luminance difference between the disk and the annulus in these cases, a chromaticity border most likely contributed little to the distinctness of the border, and it is not surprising that Hypothesis A was not supported in these conditions. Thus, as a whole, the argument of the studies by Hanada (2012, 2015) that border indistinctness between the disk and the annulus weakens the feeling of being dazzled was also supported by this study. The stimuli with the disk-annulus stimulus configuration somewhat resembled the shading pattern of an object-like a sphere. Thus, the visual system may have interpreted this configuration as an object as well as a light source. The indistinctness of the border between the disk and the annulus may have enhanced the object-like appearance, which may have inhibited the perception of luminosity and eventually led to the reduction of the feeling of being dazzled.

An increasing saturation gradient should be more natural than the other gradients, as explained in the introduction. Based on this information, Hypothesis B was the following: the feeling of being dazzled would be stronger for natural saturation gradients (i.e., increasing saturation gradients) than unnatural saturation gradients (i.e., decreasing and no-saturation gradients). However, for the disk-annulus configuration, the feeling of being dazzled was weaker for the increasing saturation gradient than for the other two saturation gradients when the luminance continuously changed from the disk to the annulus. This result contradicts Hypothesis B and suggests that the naturalness of the saturation gradient does not contribute to the feeling of being dazzled.

When the center luminance was lower than the periphery maximum luminance, the increasing saturation gradient evoked a stronger feeling of being dazzled compared with the other gradients for the disk-annulus configuration (Figure 5). This result was consistent with Hypothesis B, but it was not expected to hold in this condition because a glare image whose disk luminance is lower than the annulus maximum luminance is still unnatural. Some individuals who have observed the images in this condition have reported that the disk appears brighter and more whitish for the increasing saturation gradient than for the others. These factors may contribute to the differences in the feeling of being dazzled between the saturation gradient conditions. This discussion was based on informal observation, and further studies are required to validate or invalidate this explanation.

The results for the stimulus configuration of the star were similar to those for the disk-annulus, as seen in Figures 4 and 5. However, when the center and the periphery had high luminance (65 cd/m²), the results for the star were slightly different from those for the disk-annulus. The effect of the boundary enhancement on the feeling of being dazzled seemed to be reduced for the no-saturation gradient when the stimulus configuration was the star. The reason for this difference in the feeling of being dazzled between the two stimulus configurations is as yet inexplicable, but this result may be related to the fact that there is a blurred border between the center and the periphery for the star due to the angular luminance modulation in the periphery, even when the center luminance and the periphery maximum luminance are the same. However, the overall pattern of the results was similar for the stimulus configurations of the disk-annulus and the star. This suggests that a realistic depiction of glare with streaks is not crucial to the feeling of being dazzled.

The results for the stimulus configuration of the cross were quite different from those of the disk-annulus and the star, as shown in Figure 5. The overall effects of the periphery saturation gradient were weaker for the cross than for the disk-annulus and the star. For the 65 cd/m² center and maximum peripheral luminances, its effects for the cross was opposite for those for the other stimulus configurations. These results indicate that for the stimulus configuration of the cross, the distinctness of the central square did not enhance the feeling of being dazzled. The reason for this may be that a chromatic border between the center and the periphery would not affect the object-like or light-emitting appearance of the cross, because the cross stimulus has a relatively distinct figure against the background that causes it to look like an object irrespective of the saturation gradient condition. The significant effect of the periphery saturation gradient for the 65 cd/m² center and maximum periphery luminances was consistent with Hypothesis B and may be explained by naturalness of the increasing saturation gradient. However, as Hypothesis B was not supported for the stimulus configuration of the disk-annulus and the star, the explanation in terms of the naturalness of the color saturation gradient is implausible. We do not have any explanation for this at present. Further studies on the feeling of being dazzling evoked by the cross shape are needed.

Zavagno (1999) presented a number of visual stimuli that evoked a luminous appearance and showed that the details of the stimulus configuration do not affect the luminosity perception evoked by the luminance gradient. However, effects of the periphery saturation gradient on the feeling of being dazzled depend on the stimulus configuration. This fact suggests that luminosity may be affected by the stimulus configuration to some extent. Zavagno and Daneyko (2017) presented demonstrative figures that may support this suggestion. From the comparison of the glare effect and the phantom illumination illusion induced by similar luminance gradients with different backgrounds (black or white), it was suggested that the self-luminous appearance depends on the contours surrounding the central region. Another possibility is that the feeling of being dazzled was not necessarily determined by the perceived luminosity of visual stimuli and should be affected by many factors. Furthermore, the results of this study suggest that studies using glare images with a luminance gradient should take stimulus configurations into account, even though the stimulus configuration may not greatly affect perceived luminosity.

The glare images used in this study have been used in computer graphics, motion pictures, animation film, and artistic pictures. The results of this study also have some practical implications. To reduce the feeling of being dazzled evoked by glare stimuli with a concentric or star shape, the saturation of color should increase gradually from the center to the periphery to make the central region indistinct. To strengthen the feeling of being dazzled, the chromatic difference between the center and the periphery should be increased to underline the boundary between the center and the periphery. A blue periphery with a luminance gradient is also effective in strengthening the feeling of being dazzled (see also Hanada, 2015). The feeling of being dazzled has not been studied intensively, though it is a common experience in everyday life, and it seems to be a fundamental feeling related to visual perception. This study provides important information for the understanding of the mechanism to evoke that feeling and to control it.

However, there are certain limitations in this study. Although this study used only three stimulus configurations, there are many more possible stimulus configurations, such as, for example, the image of a light source partially occluded by objects in front of it. Moreover, an achromatic color was used for the central region in this study, but other colors may affect the effects of the periphery saturation gradients on the feeling of being dazzled. We have to clarify to what extent the findings of this study can be generalized. It should also be noted that we cannot know which of the stimulus configurations used in this study is most effective in evoking the feeling of being dazzled because different participants rated the feeling, and their judgments were rather relative than absolute. (All the stimuli were presented once to the participants before the experimental trial to help them standardize their responses.) Further studies are needed to address these issues.