Influence of luminance distribution on the ... - Semantic Scholar
Appetite 54 (2010) 363–368
Contents lists available at ScienceDirect
Appetite journal homepage: www.elsevier.com/locate/appet
Research report
Influence of luminance distribution on the appetizingly fresh appearance of cabbage Yuji Wada a,1,*, Carlos Arce-Lopera b,1, Tomohiro Masuda a, Atsushi Kimura a, Ippeita Dan a, Sho-ichi Goto c, Daisuke Tsuzuki c,d, Katsunori Okajima e,* a
National Food Research Institute, Sensory and Cognitive Food Science Laboratory, 2-1-12, Kannondai, Tsukuba, Ibaraki 305-8642, Japan Graduate School of Environment and Information Sciences, Yokohama National University, 79-7, Tokiwadai, Hodogaya-ku, Yokohama, Kanagawa 240-8501, Japan Graduate School of System and Information Engineering, University of Tsukuba, 1-1-1 Tennodai, Tsukuba, Ibaraki 305-8573, Japan d The Japan Society for the Promotion of Science, Tokyo 102-8472, Japan e Research Institute of Environment and Information Sciences, Yokohama National University, 79-7, Tokiwadai, Hodogaya-ku, Yokohama, Kanagawa 240-8501, Japan b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 28 April 2009 Received in revised form 28 December 2009 Accepted 5 January 2010
We investigated the effect that the parameters of luminance distribution in fresh food have on our visual perception of its freshness. We took pictures of the degradation over 32 h in freshness of a cabbage. We used original images, which were patches of the pictures taken at different sampling hours, and artificially generated pictures, called ‘‘matched images,’’ created by fitting the luminance histogram shape of the original image (taken at the 1st hour) to those at various freshness stages using a luminance histogram-matching algorithm. Nine participants rated the perceived freshness of the original and the matched images on a scale of degradation. As a result, we found that the participants could quantitatively estimate the degradation in freshness of the cabbage simply by looking at the presented images. Some parameters of the luminance histograms monotonically change with decreasing freshness, indicating that the freshness of cabbage can be estimated using these parameters. However, the freshness ratings for the matched images after the 8th hour of degradation had lower modification than those for the respective original images. These results suggest that the luminance distribution in the vegetable texture partly contributes to visual freshness perception but other variables, such as spatial patterns, might also be important for estimating visual freshness. ß 2010 Elsevier Ltd. All rights reserved.
Keywords: Visual freshness perception Histogram matching technique Image statistics Cabbage
Introduction In daily shopping, we quickly choose the vegetables that look fresh and appetizing without touching or tasting them. This fact suggests that humans can estimate the quality of food, such as the deterioration or the freshness of vegetables, using optical cues. However, it remains unknown how humans perform this quality assessment. In general, one of the most plausible cues for the estimation of visual freshness might be colour perception since the tri-chromatic colour vision in primates is traditionally believed to have evolved to detect ripe fruit on a dappled background of leaves (Mollon, 1989; Osorio & Vorobyev, 1996; Regan et al., 2001). However, some researchers have reported that, compared to other factors, colour appearance contributes to a lesser extent to freshness perception in some fruits and vegetables (Pe´neau, Brockhoff, Escher, & Nuessli, 2007). For example, Pe´neau et al. (2007) investigated the sensory
* Corresponding authors. E-mail addresses: [email protected] (Y. Wada), [email protected] (K. Okajima). 1 Equal contribution. 0195-6663/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.appet.2010.01.002
attributes influencing consumer perception of the freshness of strawberries and carrots. The sensory characters of products, which include not only taste and odor, but also visual parameters such as shininess (or glossiness) and colour, were evaluated by trained panels, indicating that subjective visual parameters such as bruising and shininess are the best predictors of consumer perception of freshness, while colour does not strongly contribute to freshness perception. In this study, the relationship between subjective sensory attributes and freshness was examined. However, the relationship between freshness and sensory attributes such as shininess is just beginning to be understood, and the optical or physical cues relating to freshness perception are still relatively unknown. In order to examine the relationship between some physical parameters and freshness perception, experimental approaches using psychophysical techniques are necessary. Recently, a physical property that might serve as a visual freshness cue was reported in visual science: Motoyoshi, Nishida, Sharan, and Adelson (2007) showed that image statistic values, such as the distribution characteristics of luminance, are highly correlated with the perceived glossiness and lightness of the visual texture. They demonstrated that as the luminance distribution in
Y. Wada et al. / Appetite 54 (2010) 363–368
364
the observed image became positively skewed, perceived glossiness of the image increased. Moreover, they found that observers’ visual adaptation to a positively skewed image made subsequent images look less glossy, and concluded that the human visual system can extract the luminance distribution skewness to be able to perceive the quality of a surface. Thus, we can expect that the distribution characteristics of luminance may play an important role in the mechanism of our visual freshness perception. In the present study, we used this insight and investigated whether optical features affect our visual perception of freshness in cabbages. We chose a cabbage as the sample because a fresh cabbage looks glossy and leafy vegetables like this one are highly perishable. In addition, the freshness of cabbage is generally proportional to its appetizing appearance. First, we took digital photographs of the freshness degradation process of the cabbage over 32 h. As visual stimuli, we used two different sets of images: original images and matched images. The first set was the original images consisting of 512 512 pixel patches of the original pictures. Using them, we investigated whether the freshness of cabbage would be adequately perceived with visual cues. The other set was the matched images consisting of 512 512 pixel patches of the artificial images created by modifying their luminance distribution. We used this set to clarify the effect of the luminance distribution on freshness perception leaving aside the effects of other visual attributes, such as colour and shape. Participants were required to rate the freshness of the cabbage in the images which they observed. The current study provides the first attempt to examine the effect of image statistic parameters on the perception of the visual freshness of food. Methods The psychophysical experiment consisted of two separate sessions, which were different only in the visual stimuli observed by the participants. In the first experimental session, we used the original images of the cabbage at various freshness degradation stages. In the second session, the matched images that had been generated artificially were used as the visual stimuli. Participants were asked to estimate the freshness of the cabbage in the images using a Visual Analogue Scale (VAS) in all trials of both sessions. Participants There were 9 participants, 5 females and 4 males ranging in age from 23 to 41 years old, in both experimental sessions. All of them
had normal colour vision, and normal or corrected to normal visual acuity. No experts on cooking, trading, farming vegetables, or sensory evaluation of food were included. We conducted no specific training for participants. Written informed consent was obtained after a complete explanation of the study. The study was approved by the institutional ethics committee of the National Food Research Institute. Apparatus and stimuli Apparatus The visual stimuli were presented on a 22-in. CRT monitor (Iiyama HM204DA) using ViSaGe (Cambridge Research Systems Co. Ltd.). The viewing distance between the display and the chinrest was about 57 cm. Sample We used one leaf from a fresh head of cabbage that we randomly selected from a local market on January 18th, 2008. The photographs used in the experiments were taken on the date of and the date after purchase. Original images The images used in the experiment, and which we call original images, were taken in a darkroom in which the humidity and the temperature were kept at 6% and 30 8C, respectively. A digital camera (Nikon COOLPIX P5100) was set using a tripod stand in a box designed for taking photographs (D’ CUBE J; 116 100 100 cm). Illumination was achieved with 2 floor lamps having a colour temperature of 5400 K. We took 3000 4000 pixel photos every hour for 32 h, automatically. As original-images stimuli, we used patches of 512 512 pixels of the pictures of the freshness degradation process taken at 1, 2, 3, 5, 8, 11, 15, 19, 23, 27, and 32 h (see Fig. 1a). All images were cropped at the same approximate coordinates. The purpose of this selection was to investigate whether observers could perceive freshness as a negative function of degradation time. Fig. 2 shows the most common statistics that characterize the luminance histogram distribution of the original images (a) and the matched images (b); standard deviation, skewness, and kurtosis on a normalized scale on the vertical axis, and time on the horizontal axis. This figure shows that the luminance distribution of the matched images are almost the same as those of the original images, and the luminance distribution features change, as anticipated, as a function of the degradation duration.
Fig. 1. Examples of visual stimuli. (a) Original images of 1st, 3rd, 8th, 19th, and 32nd, hours; (b) matched images of 1st, 3rd, 8th, 19th, and 32nd, hours.
Y. Wada et al. / Appetite 54 (2010) 363–368
365
Fig. 2. Normalized statistical measures of the luminance histograms of both kinds of images as functions of degradation duration. (a) Original images; (b) matched images.
Table 1 Means of colour parameters in each original image. Degradation duration (h)
1
2
3
5
8
11
15
19
23
27
32
L* a* b* Chroma Hue D*Eab between 1st hour image and others
81.72 16.86 0.40 16.87 181.37 0.00
82.49 16.72 0.22 16.72 180.74 0.80
82.81 16.71 0.40 16.71 181.38 1.10
82.91 16.76 0.20 16.76 180.69 1.21
82.68 16.84 0.20 16.84 179.34 1.13
83.23 16.78 0.17 16.78 180.57 1.53
83.71 16.43 0.42 16.43 181.46 2.03
83.85 16.48 0.46 16.49 181.61 2.16
83.66 16.48 0.04 16.48 179.86 2.02
83.88 16.63 0.11 16.63 179.62 2.23
83.34 16.95 0.44 16.95 178.53 1.82
In addition, possible effects of colour parameters were assessed. As shown in Table 1, we found only small changes in the averages of colour parameters in these patches: the colour differences (D*Eab) between the 1st hour image and the other images were smaller than 2.3 which corresponds approximately to a justnoticeable difference (Sharma, 2003). Based on this observation, we assumed that the effect of the colour difference of the stimuli on human perception was negligible. Matched images The artificially generated pictures, called matched images, were digitally created by manipulating their luminance distribution using the histogram matching technique used by Horn and Woodham (1979). Fig. 3 shows a schematic illustration of this technique. Basically, two inputs are necessary: base and target histograms. The shape of the base histogram changes to match the target’s shape. Each matched image was corrected to keep the shape, the mean luminance value, and the chromatic values of the original image. In Fig. 3 an example is shown: the base is the 1st hour original image, and the target is the luminance histogram of the 32nd hour original image. Once the luminance matching technique was performed, a new digital stimulus was created, which we called the 32nd-hour matched image. We used the original image of the 1st hour as the base and generated 10 matched images by using the luminance histograms of the other 10 original images as targets (Fig. 1b). Fig. 4a and 4b show the shifts of power spectrums of the original images and the matched images, respectively. In these figures, the vertical axis indicates frequency range, and the horizontal axis indicates degradation duration. Increasing gray intensities indicate increasing magnitudes of power spectrum as log10 of output. Comparison between the two graphs in Fig. 4 indicates that spatial frequency patterns in both image groups are quite different.
Therefore, any correlations found in the results of both experimental sessions must not depend on spatial frequency patterns. In addition, colour parameters were almost identical for all matched images as shown in Table 2. Procedure Both experimental sessions adopted 1 11 paradigms including the degradation duration (1st, 2nd, 3rd, 5th, 8th, 11th, 15th, 19th, 23rd, 27th, and 32nd hours). The participants’ heads were fixed to a chin rest about 57 cm from the screen. Participants binocularly observed the presented stimuli in a dark room after a dark adaptation period of 10 min. To rate the perceived freshness, we used a 10 cm visual analogue scale with end anchors, printed on paper sheets. At the beginning of the experiment, two reference images were presented on the screen of the CRT monitor: one with a value of 2 on the freshness scale corresponding to the patch extracted from the digital photograph at the 32nd hour (old) and the other with a value of 8 on the freshness scale corresponding to the patch extracted from the photograph at the 1st hour (new). Participants were asked to rate the perceived freshness as a negative function of degradation time of the cabbage leaf in the images, using this scale, for 66 trials (11 images 3 times 2 sessions). Participants were allowed to check any point on the scale even if the checked point was not exactly on a tick mark. In both sessions, the images were presented in random order. Analysis of data For each participant, we calculated the mean of the rating (length on the scale) for each condition as her or his perceived freshness. In order to examine the relationship between the
366
Y. Wada et al. / Appetite 54 (2010) 363–368
Fig. 3. Schematic illustrations of the histogram matching technique used when generating the matched images. In this example, the base is the 1st hour original image (upper left). The target is the luminance histogram of the 32nd hour original image (bottom left). Once the luminance matching technique is performed a new, digital stimuli is created (right).
Fig. 4. Power spectrum of the spatial frequency of both kinds of images as a function of frequency range and degradation duration. Increasing gray intensities indicate increasing magnitude of power spectrum as log10 of output. (a) Original images; (b) matched images.
freshness degradation duration and the perceived freshness, a regression analysis of log10 degradation duration versus log10 estimated freshness was conducted using the analogue scale for both sessions.
We used a two-way repeated measure of analysis of variance (ANOVA) with the two kinds of images and the degradation duration (10 degrees) on the means of the freshness ratings for each condition. We excluded the data under the 1st hour condition
Y. Wada et al. / Appetite 54 (2010) 363–368
367
Table 2 Means of colour parameters in each matched image. Degradation duration (h)
1
2
3
5
8
11
15
19
23
27
32
L* a* b* Chroma Hue D*Eab between 1st hour image and others
81.72 16.86 0.40 16.87 181.37 0.00
81.68 16.83 0.45 16.84 181.52 0.06
81.50 16.84 0.41 16.84 181.39 0.22
81.57 16.79 0.50 16.80 181.72 0.19
81.36 16.81 0.42 16.81 181.44 0.37
81.40 16.79 0.47 16.79 181.60 0.34
81.30 16.80 0.42 16.81 181.42 0.43
81.50 16.75 0.55 16.76 181.88 0.29
81.31 16.79 0.44 16.80 181.49 0.42
81.54 16.74 0.57 16.75 181.94 0.27
81.54 16.75 0.57 16.75 181.95 0.27
for this analysis because both the original and matched images were identical. In order to clarify the relationship between the luminance histogram parameters and the perceived freshness, we calculated their Pearson’s product moment correlation coefficient for each session. Results Regression analysis of means across participants The results are shown in Fig. 5. In our regression analysis for the original images, the intercept and the slope of the mean regression line were 4.383 and 0.355, respectively (R2 = 0.964, F(1, 9) = 242.513, p < 0.01). For the matched images, the intercept and the slope of the mean regression line were 4.102 and 0.066, respectively (R2 = 0.424, F(1, 9) = 6.615, p < 0.05).
Table 3 F values of significant results of post hoc analysis between the rating for original images and that for matched images. Degradation duration (h)
F(1, 80)
p
8 11 15 19 23 27 32
5.22 4.50 20.57 27.15 19.64 29.64 27.02
Contents lists available at ScienceDirect
Appetite journal homepage: www.elsevier.com/locate/appet
Research report
Influence of luminance distribution on the appetizingly fresh appearance of cabbage Yuji Wada a,1,*, Carlos Arce-Lopera b,1, Tomohiro Masuda a, Atsushi Kimura a, Ippeita Dan a, Sho-ichi Goto c, Daisuke Tsuzuki c,d, Katsunori Okajima e,* a
National Food Research Institute, Sensory and Cognitive Food Science Laboratory, 2-1-12, Kannondai, Tsukuba, Ibaraki 305-8642, Japan Graduate School of Environment and Information Sciences, Yokohama National University, 79-7, Tokiwadai, Hodogaya-ku, Yokohama, Kanagawa 240-8501, Japan Graduate School of System and Information Engineering, University of Tsukuba, 1-1-1 Tennodai, Tsukuba, Ibaraki 305-8573, Japan d The Japan Society for the Promotion of Science, Tokyo 102-8472, Japan e Research Institute of Environment and Information Sciences, Yokohama National University, 79-7, Tokiwadai, Hodogaya-ku, Yokohama, Kanagawa 240-8501, Japan b c
A R T I C L E I N F O
A B S T R A C T
Article history: Received 28 April 2009 Received in revised form 28 December 2009 Accepted 5 January 2010
We investigated the effect that the parameters of luminance distribution in fresh food have on our visual perception of its freshness. We took pictures of the degradation over 32 h in freshness of a cabbage. We used original images, which were patches of the pictures taken at different sampling hours, and artificially generated pictures, called ‘‘matched images,’’ created by fitting the luminance histogram shape of the original image (taken at the 1st hour) to those at various freshness stages using a luminance histogram-matching algorithm. Nine participants rated the perceived freshness of the original and the matched images on a scale of degradation. As a result, we found that the participants could quantitatively estimate the degradation in freshness of the cabbage simply by looking at the presented images. Some parameters of the luminance histograms monotonically change with decreasing freshness, indicating that the freshness of cabbage can be estimated using these parameters. However, the freshness ratings for the matched images after the 8th hour of degradation had lower modification than those for the respective original images. These results suggest that the luminance distribution in the vegetable texture partly contributes to visual freshness perception but other variables, such as spatial patterns, might also be important for estimating visual freshness. ß 2010 Elsevier Ltd. All rights reserved.
Keywords: Visual freshness perception Histogram matching technique Image statistics Cabbage
Introduction In daily shopping, we quickly choose the vegetables that look fresh and appetizing without touching or tasting them. This fact suggests that humans can estimate the quality of food, such as the deterioration or the freshness of vegetables, using optical cues. However, it remains unknown how humans perform this quality assessment. In general, one of the most plausible cues for the estimation of visual freshness might be colour perception since the tri-chromatic colour vision in primates is traditionally believed to have evolved to detect ripe fruit on a dappled background of leaves (Mollon, 1989; Osorio & Vorobyev, 1996; Regan et al., 2001). However, some researchers have reported that, compared to other factors, colour appearance contributes to a lesser extent to freshness perception in some fruits and vegetables (Pe´neau, Brockhoff, Escher, & Nuessli, 2007). For example, Pe´neau et al. (2007) investigated the sensory
* Corresponding authors. E-mail addresses: [email protected] (Y. Wada), [email protected] (K. Okajima). 1 Equal contribution. 0195-6663/$ – see front matter ß 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.appet.2010.01.002
attributes influencing consumer perception of the freshness of strawberries and carrots. The sensory characters of products, which include not only taste and odor, but also visual parameters such as shininess (or glossiness) and colour, were evaluated by trained panels, indicating that subjective visual parameters such as bruising and shininess are the best predictors of consumer perception of freshness, while colour does not strongly contribute to freshness perception. In this study, the relationship between subjective sensory attributes and freshness was examined. However, the relationship between freshness and sensory attributes such as shininess is just beginning to be understood, and the optical or physical cues relating to freshness perception are still relatively unknown. In order to examine the relationship between some physical parameters and freshness perception, experimental approaches using psychophysical techniques are necessary. Recently, a physical property that might serve as a visual freshness cue was reported in visual science: Motoyoshi, Nishida, Sharan, and Adelson (2007) showed that image statistic values, such as the distribution characteristics of luminance, are highly correlated with the perceived glossiness and lightness of the visual texture. They demonstrated that as the luminance distribution in
Y. Wada et al. / Appetite 54 (2010) 363–368
364
the observed image became positively skewed, perceived glossiness of the image increased. Moreover, they found that observers’ visual adaptation to a positively skewed image made subsequent images look less glossy, and concluded that the human visual system can extract the luminance distribution skewness to be able to perceive the quality of a surface. Thus, we can expect that the distribution characteristics of luminance may play an important role in the mechanism of our visual freshness perception. In the present study, we used this insight and investigated whether optical features affect our visual perception of freshness in cabbages. We chose a cabbage as the sample because a fresh cabbage looks glossy and leafy vegetables like this one are highly perishable. In addition, the freshness of cabbage is generally proportional to its appetizing appearance. First, we took digital photographs of the freshness degradation process of the cabbage over 32 h. As visual stimuli, we used two different sets of images: original images and matched images. The first set was the original images consisting of 512 512 pixel patches of the original pictures. Using them, we investigated whether the freshness of cabbage would be adequately perceived with visual cues. The other set was the matched images consisting of 512 512 pixel patches of the artificial images created by modifying their luminance distribution. We used this set to clarify the effect of the luminance distribution on freshness perception leaving aside the effects of other visual attributes, such as colour and shape. Participants were required to rate the freshness of the cabbage in the images which they observed. The current study provides the first attempt to examine the effect of image statistic parameters on the perception of the visual freshness of food. Methods The psychophysical experiment consisted of two separate sessions, which were different only in the visual stimuli observed by the participants. In the first experimental session, we used the original images of the cabbage at various freshness degradation stages. In the second session, the matched images that had been generated artificially were used as the visual stimuli. Participants were asked to estimate the freshness of the cabbage in the images using a Visual Analogue Scale (VAS) in all trials of both sessions. Participants There were 9 participants, 5 females and 4 males ranging in age from 23 to 41 years old, in both experimental sessions. All of them
had normal colour vision, and normal or corrected to normal visual acuity. No experts on cooking, trading, farming vegetables, or sensory evaluation of food were included. We conducted no specific training for participants. Written informed consent was obtained after a complete explanation of the study. The study was approved by the institutional ethics committee of the National Food Research Institute. Apparatus and stimuli Apparatus The visual stimuli were presented on a 22-in. CRT monitor (Iiyama HM204DA) using ViSaGe (Cambridge Research Systems Co. Ltd.). The viewing distance between the display and the chinrest was about 57 cm. Sample We used one leaf from a fresh head of cabbage that we randomly selected from a local market on January 18th, 2008. The photographs used in the experiments were taken on the date of and the date after purchase. Original images The images used in the experiment, and which we call original images, were taken in a darkroom in which the humidity and the temperature were kept at 6% and 30 8C, respectively. A digital camera (Nikon COOLPIX P5100) was set using a tripod stand in a box designed for taking photographs (D’ CUBE J; 116 100 100 cm). Illumination was achieved with 2 floor lamps having a colour temperature of 5400 K. We took 3000 4000 pixel photos every hour for 32 h, automatically. As original-images stimuli, we used patches of 512 512 pixels of the pictures of the freshness degradation process taken at 1, 2, 3, 5, 8, 11, 15, 19, 23, 27, and 32 h (see Fig. 1a). All images were cropped at the same approximate coordinates. The purpose of this selection was to investigate whether observers could perceive freshness as a negative function of degradation time. Fig. 2 shows the most common statistics that characterize the luminance histogram distribution of the original images (a) and the matched images (b); standard deviation, skewness, and kurtosis on a normalized scale on the vertical axis, and time on the horizontal axis. This figure shows that the luminance distribution of the matched images are almost the same as those of the original images, and the luminance distribution features change, as anticipated, as a function of the degradation duration.
Fig. 1. Examples of visual stimuli. (a) Original images of 1st, 3rd, 8th, 19th, and 32nd, hours; (b) matched images of 1st, 3rd, 8th, 19th, and 32nd, hours.
Y. Wada et al. / Appetite 54 (2010) 363–368
365
Fig. 2. Normalized statistical measures of the luminance histograms of both kinds of images as functions of degradation duration. (a) Original images; (b) matched images.
Table 1 Means of colour parameters in each original image. Degradation duration (h)
1
2
3
5
8
11
15
19
23
27
32
L* a* b* Chroma Hue D*Eab between 1st hour image and others
81.72 16.86 0.40 16.87 181.37 0.00
82.49 16.72 0.22 16.72 180.74 0.80
82.81 16.71 0.40 16.71 181.38 1.10
82.91 16.76 0.20 16.76 180.69 1.21
82.68 16.84 0.20 16.84 179.34 1.13
83.23 16.78 0.17 16.78 180.57 1.53
83.71 16.43 0.42 16.43 181.46 2.03
83.85 16.48 0.46 16.49 181.61 2.16
83.66 16.48 0.04 16.48 179.86 2.02
83.88 16.63 0.11 16.63 179.62 2.23
83.34 16.95 0.44 16.95 178.53 1.82
In addition, possible effects of colour parameters were assessed. As shown in Table 1, we found only small changes in the averages of colour parameters in these patches: the colour differences (D*Eab) between the 1st hour image and the other images were smaller than 2.3 which corresponds approximately to a justnoticeable difference (Sharma, 2003). Based on this observation, we assumed that the effect of the colour difference of the stimuli on human perception was negligible. Matched images The artificially generated pictures, called matched images, were digitally created by manipulating their luminance distribution using the histogram matching technique used by Horn and Woodham (1979). Fig. 3 shows a schematic illustration of this technique. Basically, two inputs are necessary: base and target histograms. The shape of the base histogram changes to match the target’s shape. Each matched image was corrected to keep the shape, the mean luminance value, and the chromatic values of the original image. In Fig. 3 an example is shown: the base is the 1st hour original image, and the target is the luminance histogram of the 32nd hour original image. Once the luminance matching technique was performed, a new digital stimulus was created, which we called the 32nd-hour matched image. We used the original image of the 1st hour as the base and generated 10 matched images by using the luminance histograms of the other 10 original images as targets (Fig. 1b). Fig. 4a and 4b show the shifts of power spectrums of the original images and the matched images, respectively. In these figures, the vertical axis indicates frequency range, and the horizontal axis indicates degradation duration. Increasing gray intensities indicate increasing magnitudes of power spectrum as log10 of output. Comparison between the two graphs in Fig. 4 indicates that spatial frequency patterns in both image groups are quite different.
Therefore, any correlations found in the results of both experimental sessions must not depend on spatial frequency patterns. In addition, colour parameters were almost identical for all matched images as shown in Table 2. Procedure Both experimental sessions adopted 1 11 paradigms including the degradation duration (1st, 2nd, 3rd, 5th, 8th, 11th, 15th, 19th, 23rd, 27th, and 32nd hours). The participants’ heads were fixed to a chin rest about 57 cm from the screen. Participants binocularly observed the presented stimuli in a dark room after a dark adaptation period of 10 min. To rate the perceived freshness, we used a 10 cm visual analogue scale with end anchors, printed on paper sheets. At the beginning of the experiment, two reference images were presented on the screen of the CRT monitor: one with a value of 2 on the freshness scale corresponding to the patch extracted from the digital photograph at the 32nd hour (old) and the other with a value of 8 on the freshness scale corresponding to the patch extracted from the photograph at the 1st hour (new). Participants were asked to rate the perceived freshness as a negative function of degradation time of the cabbage leaf in the images, using this scale, for 66 trials (11 images 3 times 2 sessions). Participants were allowed to check any point on the scale even if the checked point was not exactly on a tick mark. In both sessions, the images were presented in random order. Analysis of data For each participant, we calculated the mean of the rating (length on the scale) for each condition as her or his perceived freshness. In order to examine the relationship between the
366
Y. Wada et al. / Appetite 54 (2010) 363–368
Fig. 3. Schematic illustrations of the histogram matching technique used when generating the matched images. In this example, the base is the 1st hour original image (upper left). The target is the luminance histogram of the 32nd hour original image (bottom left). Once the luminance matching technique is performed a new, digital stimuli is created (right).
Fig. 4. Power spectrum of the spatial frequency of both kinds of images as a function of frequency range and degradation duration. Increasing gray intensities indicate increasing magnitude of power spectrum as log10 of output. (a) Original images; (b) matched images.
freshness degradation duration and the perceived freshness, a regression analysis of log10 degradation duration versus log10 estimated freshness was conducted using the analogue scale for both sessions.
We used a two-way repeated measure of analysis of variance (ANOVA) with the two kinds of images and the degradation duration (10 degrees) on the means of the freshness ratings for each condition. We excluded the data under the 1st hour condition
Y. Wada et al. / Appetite 54 (2010) 363–368
367
Table 2 Means of colour parameters in each matched image. Degradation duration (h)
1
2
3
5
8
11
15
19
23
27
32
L* a* b* Chroma Hue D*Eab between 1st hour image and others
81.72 16.86 0.40 16.87 181.37 0.00
81.68 16.83 0.45 16.84 181.52 0.06
81.50 16.84 0.41 16.84 181.39 0.22
81.57 16.79 0.50 16.80 181.72 0.19
81.36 16.81 0.42 16.81 181.44 0.37
81.40 16.79 0.47 16.79 181.60 0.34
81.30 16.80 0.42 16.81 181.42 0.43
81.50 16.75 0.55 16.76 181.88 0.29
81.31 16.79 0.44 16.80 181.49 0.42
81.54 16.74 0.57 16.75 181.94 0.27
81.54 16.75 0.57 16.75 181.95 0.27
for this analysis because both the original and matched images were identical. In order to clarify the relationship between the luminance histogram parameters and the perceived freshness, we calculated their Pearson’s product moment correlation coefficient for each session. Results Regression analysis of means across participants The results are shown in Fig. 5. In our regression analysis for the original images, the intercept and the slope of the mean regression line were 4.383 and 0.355, respectively (R2 = 0.964, F(1, 9) = 242.513, p < 0.01). For the matched images, the intercept and the slope of the mean regression line were 4.102 and 0.066, respectively (R2 = 0.424, F(1, 9) = 6.615, p < 0.05).
Table 3 F values of significant results of post hoc analysis between the rating for original images and that for matched images. Degradation duration (h)
F(1, 80)
p
8 11 15 19 23 27 32
5.22 4.50 20.57 27.15 19.64 29.64 27.02
