Abstract
In order to obtain the physiological and psychological indicators of the visual complexity of art images from the perspective of visual cognition, this study explored the relationship between eye-tracking metrics and the psychological factors. The study invited 16 participants (8 females, age range 23.81 ± 0.98) to participate in the experiment. In this study, eye-tracking experiments and a questionnaire of psychological factors affecting visual complexity were conducted. The results show that there is a significant relationship between the fixation length, first fixation time and visual complexity. Image with the complexity score interval [74, 100] has a high mental workload on visual processing. There is a significant linear relationship between the fixation count and visual complexity. In addition, the analysis of the psychological scale shows that psychological factors have a positive significant correlation with visual complexity. The participants show sensitivity to the factor of color, texture, and cognitive on visual complexity, but were insensitive to shape factors.
Similar content being viewed by others
Keywords
1 Introduction
A series of researches such as image processing and computer vision have been developed. As a basic aspect of visual cognition, visual complexity has important research value in the fields of cognitive psychology.
Image complexity generally refers to the internal complexity of the image. The definition of image complexity is different in various academic sectors. Previous research started with the concept of complexity in composition theory, dividing the complexity of images into global complexity and local complexity, and dividing complexity features into color, texture, and shape [1]. Rump [2] recommends the use of specific dimensions in the study of visual complexity. Kreitler et al. [3] also believe that visual complexity is a multidimensional concept of 5 dimensions. According to Roberts MN [4], visual complexity can be explained from 7 dimensions, respectively are element incomprehensibility, element confusion, number of elements, element type, asymmetry, color diversity and 3D appearance.
Image complexity can be used for visual cognition and emotion evaluation research [5]. People have acquired experience in the natural environment, then transfer them to the man-made environment [6]. Human visual cognition is similar to other organisms and tends to process only a part of sensory input selectively [7]. Yanqin et al. [8] evaluated the complexity of gray-scale images based on the image complexity evaluation method of BP neural network, and verified that the algorithm conforms to human visual cognition. Hao et al. [9] proposed that the image features that affect human visual complexity mainly include 3 aspects: the number of colors, the uniformity of the spatial distribution of colors, and the number of objects in the image.
In the research on the visual complexity of art images, Guo et al. [10] proposed a framework for evaluating the visual complexity of paintings, but did not involve the cognitive measurement of visual complexity. In order to obtain the physiological and psychological indicators of the visual complexity of art images, this research explores the relationship between eye-tracking metrics, psychological factors and visual complexity through a method based on cognitive psychology.
2 Method
2.1 Stimuli
In order to achieve idealized results that meet the standard of visual complexity, we chose the open source image data set SAVOIAS provided by Elham et al. [11] as stimuli. The data set contains 1420 images in 7 categories, and the scoring method is to obtain more than 37,000 image labels from 1687 contributors, then converted the paired scores into absolute scores of [0, 100], which have high credibility on visual complexity. The stimuli are selected from the art category in the data set SAVOIAS, with a total of 420 images. All images are verified to conform to the normal distribution by using IBM SPSS Statistics R26.0.
2.2 Participants
The experiment recruited 16 college students from Shanghai Jiao Tong University. The male to female ratio was 1:1, the average age was 23.81 (SD = 0.98), the vision of all participants is normal or normal after correction, and the cognitive level are normal.
2.3 Eye-Tracking Task
The experiment equipment was Tobii T60 eye tracker, screen resolution was 1280 pixels by 1024 pixels, sampling rate was 60 Hz, and Tobii Studio 1.7.3 software was used for programming, generating area of interest (AOI) and recording experimental data. The experiment was proceeded in a quiet laboratory with suitable indoor light.
Participants signed the experiment assignment and kept 60 cm away from the eye tracker to complete the calibration. The experiment started with the experiment guide. After confirmed, participants could hit the space bar to start the eye-tracking task. The experiment program presents 420 art images in random order, each image is presented for 3 s, and there is a 1 s blank interval between each 2 images. Participants need to try to understand the content of the image while viewing the image in order to achieve the purpose of understanding the complexity of the image.
2.4 Psychological Scaling
Participants were required to fill in a questionnaire on factors affecting visual complexity after completing the eye-tracking task. According to the measurement of image complexity and the research progress of cognitive psychology, the factors are divided into 4 dimensions, which are the following 14 factors: color quantity, color harmony, color contrast, texture roughness, unit texture quantity, shape regularity, edge clarity, shape continuity, familiarity, aesthetic value, degree of abstraction, number of semantic objects, semantic intelligibility and style specificity. Participants need to evaluate on 7-level Likert scale according to the degree of influence of each factor on image complexity.
3 Result
The research results of this article are divided into two aspects, respectively are physiological data analysis and psychological data analysis. First, we analyzed the relationship between the metrics of eye-tracking behavior and visual complexity. In previous studies [12], people with high cognitive demand had longer fixation time and more fixation counts, both of which reflected longer cognitive processing and higher mental load level. Some studies have also found that the average and total time of fixation will increase mental workload [13].
After excluding invalid eye-tracking data, we divided 420 stimuli into 7 groups according to their complexity score and analyzed significance level between fixation length, first fixation time, fixation count and visual complexity.
The study performed Kruskal-Wallis test on fixation time and visual complexity (Fig. 1). Among them, there are significant differences between group 3 and group 7 (Adj. Sig. a = 0.009), group 4 and group 7 (Adj. Sig. a = 0.012), group 5 and group 7 (Adj. Sig. a = 0.016).
The Kruskal-Wallis test of first fixation time and visual complexity showed that there were significant differences between group 3 and group 7 (Adj. Sig. a = 0.001), group 4 and group 7 (Adj. Sig. a = 0.002), and group 6 and group 7 (Adj. Sig. a = 0.005). We drew a pairwise comparison map as shown in Fig. 2.
A One-way ANOVA was conducted between participants to compare the effects on the fixation count of complexity images (Table 1). Under the condition of 7 groups, [F (1,6) = 2.743, p = 0.012]. The post-hoc comparison from the Tamhane test showed that group 2 was significantly different from group 4 (SD = 0.130, p = 0.046) and group 7 (SD = 0.132, p = 0.006). After polynomial analysis, Linear Term (p = 0.004) shows that there is a significant linear relationship between the number of fixations and visual complexity.
To further understanding of the influence of participants’ subjective psychological assessment on image complexity, 16 effective questionnaires were analyzed (Fig. 3). Through the One-sample T-Test, color (M = 4.750, P = 0.001), texture (M = 4.969, P = 0.002) and cognitive (M = 4.688, P = 0.005) factors have a positive correlation with the degree of visual complexity perception The shape factor (M = 4.458, P = 0.079) has no relevance to the perception of image complexity.
Through Pearson correlation analysis, we found that both the shape factor (r = 0.676) and the cognitive factor (r = 0.852) have a significant positive correlation with the level of visual complexity (Table 2). That is, high visual complexity reflects high level of shape factors and cognitive factors and vice versa. It is worth noting that the positive correlation between shape factors and cognitive factors and the level of visual complexity does not mean that the two factors have significant impact on visual complexity. In addition, the texture factor (r = −0.212) has a negative correlation with the color and cognitive factors (r = −0.075).
4 Discussion
This study analyzed the significance level between fixation length, first fixation time, and fixation count and visual complexity. The analysis of fixation length showed that there were significant differences between 3rd, 4th, 5th group and the 7th group. According to the visual complexity level, the complexity score interval of the 7th group is [74, 100]. Image in group 7 are mostly depicted grand scenes and incomprehensible semantics, which have a high load on visual processing. The complexity score interval of groups 3, 4, and 5 is [31, 61], and the pictures are dominated by a moderate number of salient objects and easier-to-understand semantics. It can be seen that the fixation length can reflect the significance between the image with the highest complexity and the image with moderate complexity. The visual workload is significantly improved when viewing the most complex image.
The first fixation time refers to the duration of the first fixation point in area of interest (AOI). There are significant differences between the first fixation time of groups 3, 4, 6 and group 7. Whether the AOI region of the image is significant can affect the first fixation time of participants. Similar to the fixation length, fixation counts also reflects the participant’s cognitive processing and psychological load level. There were significant differences between the group 2, 4, 6 and 7. Among them, the significance of the group 2 and group 7 is 0.006, indicating that fixation count is positively correlated with complexity. From the average graph (Fig. 4), fixation count and visual complexity present a relatively good positive linear relationship.
In addition to eye-tracking data, the results of psychological scale survey showed that the participants are sensitive to the effects of color, texture and cognitive factors on visual complexity, but are not sensitive to shape factors, reflecting that the cognition process of shape in human visual system is very proficient. When the human visual system recognizes shapes, it mainly recognizes objects by first recognizing the salient parts [14]. People can make quick and accurate judgments of objects through the prominent visual parts.
In addition, the positive correlation between shape factors, cognitive factors and the level of visual complexity implicates that from the perspective of visual complexity, more attention should be paid to the level of image shape features and human psychological characteristics. Previous studies on visual complexity started from human psychological factors and explored whether the image complexity algorithm meets human psychological expectations. This research has verified the positive and significant correlation of psychological factors to the level of visual complexity through correlation tests. Follow-up research will continue to study the significance of the influence of various psychological factors on the perception of visual complexity.
From the perspective of complexity science, this study explored the level of correlation between eye-tracking metrics, physiological indicators and visual complexity. The measured feature indicators pertain to the basic features of the image. The differences in the perception of image complexity by high-level features such as semantics and style have not been analyzed. Future research will collect more data on significant influencing factors, establish a regression model between eye-tracking behavior and visual complexity, calculate a wider range of art images, and evaluate the complexity of art works from the perspective of recognition.
References
Bing, Z., Yuxia, L., Xinxin, Y., Yang, L.: Review of research on image complexity. J. Comput. Sci. 45(09), 37–44 (2018). (in Chinese)
Rump, E.E.: Is there a general factor of preference for complexity? Percept. Psychophys. 3, 346–348 (1968)
Kreitler, S., Zigler, E., Kreitler, H.: The complexity of complexity. Hum. Dev. 17, 54–73 (1974)
Roberts, M.N.: Complexity and Aesthetic Preference for Diverse Visual Stimuli. Universitat de les Illes Balears, Spain (2007)
Xiaoying, G., Wenshu, L., et al.: Computational evaluation methods of visual complexity perception for images. J. Acta Electronica Sinica 48(446(04)), 197–204 (2020). (in Chinese)
Bennett, M.R.: History of Cognitive Neuroscience. Wiley-Blackwell, Hoboken (2008)
Tirin, M., Marc, Z.: Neural mechanisms of selective visual attention. Annu. Rev. Psychol. 68(1), 47–72 (2017)
Yanqin, C., Jin, D., Yong, Z., et al.: Research on the image complexity based on texture features. J. Chin. J. Opt. 03, 99–106 (2015). (in Chinese)
Hao, W., Jin, D., Xuehui, H., Bo, X.: Research on image complexity evaluation method based on color information. In: Proceedings of SPIE, vol. 10605. LIDAR Imaging Detection and Target Recognition, 106051Q (2017)
Guo, X., Kurita, T., Asano, C.M., Asano, A.: Visual complexity assessment of painting images. In: 20th IEEE International Conference on Image Processing (ICIP), pp. 388–392. IEEE (2013)
Elham, S., Mona, J., Margrit, B.: Visual complexity analysis using deep intermediate-layer features. Comput. Vis. Image Underst. 195, (2020). ISSN 1077-3142
Di, W., Yuntao, G., Danmin, M.: Using an eye tracker to measure information processing according to need for cognition level. Soc. Behav. Person.: Int. J. 46(11), 1869–1880 (2018)
Ellis, K.K.E.: Eye tracking metrics for workload estimation in flight deck operations (2009)
Hoffman, D., Signh, M.: Salience of visual parts. Cognition 63, 29–78 (1997)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2021 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this paper
Cite this paper
Hu, R., Weng, M., Zhang, L., Li, X. (2021). Art Image Complexity Measurement Based on Visual Cognition: Evidence from Eye-Tracking Metrics. In: Ayaz, H., Asgher, U., Paletta, L. (eds) Advances in Neuroergonomics and Cognitive Engineering. AHFE 2021. Lecture Notes in Networks and Systems, vol 259. Springer, Cham. https://doi.org/10.1007/978-3-030-80285-1_16
Download citation
DOI: https://doi.org/10.1007/978-3-030-80285-1_16
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-80284-4
Online ISBN: 978-3-030-80285-1
eBook Packages: EngineeringEngineering (R0)