Abstract
The amount of visual data available on the Web is growing explosively and it is becoming increasingly important to explore methods for automatically estimating the quality of this content in a manner that is consistent with the aesthetic perceptions of humans. The key to this challenging problem is to design an appropriate set of features to extract the aesthetic properties from content. Most previous studies designed a set of aesthetic features based on photographic criteria, which were unavoidably limited to specific examples and they lacked an interpretation based on the mechanism of human aesthetic perception. According to psychological theory, visual complexity is an important property of the stimuli, because it directly influences the viewer’s arousal level, which is believed to be closely related to aesthetic perception. In this study, we propose an alternative set of features for aesthetic estimation based on a visual complexity principle. We extracted the visual complexity properties from an input image in terms of their composition, shape, and distribution. In addition, we demonstrated that the proposed features are consistent with human perception on the complexity in our visual complexity dataset. Next, we employed these features for photo-aesthetic quality estimation using a large-scale dataset. Various experiments were conducted under different conditions and comparisons with state-of-the-art methods shows that the proposed visual complexity feature outperforms photography rule-based features and even better than deep features.
Similar content being viewed by others
Notes
Prediction results webpage link: https://www.hal.t.u-tokyo.ac.jp/~sun1101/
References
Akalin A, Yildirim K, Wilson C, Kilicoglu O (2009) Architecture and engineering students’ evaluations of house faċades: preference, complexity and impressiveness. J Environ Psychol 29(1):124–132
Arbelaez P, Maire M, Fowlkes C, Malik J (2011) Contour detection and hierarchical image segmentation. IEEE Trans Pattern Anal Mach Intell 33(5):898–916
Babin BJ, Attaway JS (2000) Atmospheric affect as a tool for creating value and gaining share of customer. J Bus Res 49(2):91–99
Berlyne D (1971) Aesthetics and psychobiology. Appleton-Century-Crofts, New York
Berlyne DE (1974) Studies in the new experimental aesthetics: Steps toward an objective psychology of aesthetic appreciation Hemisphere
Bhattacharya S, Sukthankar R, Shah M (2010) A framework for photo-quality assessment and enhancement based on visual aesthetics Proceedings of the international conference on multimedia, pp 271–280
Birkhoff GD (1933) Aesthetic measure. Cambridge Mass
Campbell A, Ciesielksi V, Qin A (2015) Feature discovery by deep learning for aesthetic analysis of evolved abstract images Evolutionary and biologically inspired music, sound, art and design. Springer, pp 27–38
Chipman SF (1977) Complexity and structure in visual patterns. J Exp Psychol Gen 106(3):269
Dalal N, Triggs B (2005) Histograms of oriented gradients for human detection 2005 IEEE Computer society conference on computer vision and pattern recognition (CVPR’05), vol 1, pp 886–893
Datta R, Joshi D, Li J, Wang JZ (2006) Studying aesthetics in photographic images using a computational approach Computer vision–ECCV 2006. Springer, pp 288–301
Datta R, Li J, Wang JZ (2008) Algorithmic inferencing of aesthetics and emotion in natural images: an exposition Image processing, 2008. ICIP 2008. 15th IEEE international conference on, pp 105–108
Dhar S, Ordonez V, Berg TL (2011) High level describable attributes for predicting aesthetics and interestingness Computer vision and pattern recognition (CVPR), 2011 IEEE conference on, pp 1657–1664
Donderi DC (2006) Visual complexity: a review. Psychol Bull 132(1):73
Eysenck HJ (1942) The experimental study of the’good gestalt’—a new approach. Psychol Rev 49(4):344
Friedenberg J, Bertamini M (2015) Aesthetic preference for polygon shape. Empir Stud Arts 33(2):144–160
García M, Badre AN, Stasko JT (1994) Development and validation of icons varying in their abstractness. Interact Comput 6(2):191–211
Guo X, Asano CM, Asano A, Kurita T, Li L (2012) Analysis of texture characteristics associated with visual complexity perception. Opt Rev 19(5):306–314
He XC, Yung NH (2004) Curvature scale space corner detector with adaptive threshold and dynamic region of support Pattern recognition, 2004. ICPR 2004. Proceedings of the 17th international conference on, vol 2, pp 791–794
Heaps C, Handel S (1999) Similarity and features of natural textures. J Exp Psychol Hum Percept Perform 25(2):299
Huhmann BA (2003) Visual complexity in banner ads: The role of color, photography, and animation. Vis Commun Q 10(3):10–17
Ichikawa S (1985) Quantitative and structural factors in the judgment of pattern complexity. Percept Psychophys 38(2):101–109
Imamoglu Ç (2000) Complexity, liking and familiarity: Architecture and non-architecture turkish students’assessments of traditional and modern house facades. J Environ Psychol 20(1):5–16
Katz BF (2002) What makes a polygon pleasing? Empir Stud Arts 20(1):1–19
Ke Y, Tang X, Jing F (2006) The design of high-level features for photo quality assessment Computer vision and pattern recognition, 2006 IEEE computer society conference on, vol 1, pp 419–426
Krizhevsky A, Sutskever I, Hinton GE (2012) Imagenet classification with deep convolutional neural networks Advances in neural information processing systems, pp 1097–1105
Leder H, Belke B, Oeberst A, Augustin D (2004) A model of aesthetic appreciation and aesthetic judgments. Br J Psychol 95(4):489–508
Li C, Loui AC, Chen T (2010) Towards aesthetics: a photo quality assessment and photo selection system Proceedings of the international conference on multimedia, pp 827–830
Lihua G, Fudi L (2015). arXiv:1505.05225
Lo KY, Liu KH, Chen CS (2012) Assessment of photo aesthetics with efficiency Pattern recognition (ICPR), 2012 21st international conference on, pp 2186–2189
Lu P, Peng X, Li R, Wang X (2015) Towards aesthetics of image: a bayesian framework for color harmony modeling. Signal Processing: Image Communication
Lu X, Lin Z, Jin H, Yang J, Wang JZ (2014) Rapid: Rating pictorial aesthetics using deep learning Proceedings of the ACM international conference on multimedia, pp 457–466
Lu X, Lin Z, Shen X, Mech R, Wang JZ (2015) Deep multi-patch aggregation network for image style, aesthetics, and quality estimation Proceedings of the IEEE international conference on computer vision, pp 990–998
Luo Y, Tang X (2008) Photo and video quality evaluation: Focusing on the subject Computer vision–ECCV 2008. Springer, pp 386–399
Luo W, Wang X, Tang X (2011) Content-based photo quality assessment Computer vision (ICCV), 2011 IEEE international conference on, pp 2206–2213
Mai L, Jin H, Liu F (2016) Composition-preserving deep photo aesthetics assessment Proceedings of the IEEE conference on computer vision and pattern recognition, pp 497–506
Mallon B, Redies C, Hayn-Leichsenring GU (2014) Beauty in abstract paintings: perceptual contrast and statistical properties. Frontiers in human neuroscience 8
Martín H JA, Santos M, de Lope J (2010) Orthogonal variant moments features in image analysis. Inf Sci 180(6):846–860
Marchesotti L, Perronnin F, Larlus D, Csurka G (2011) Assessing the aesthetic quality of photographs using generic image descriptors Computer vision (ICCV), 2011 IEEE international conference on, pp 1784–1791
Munsinger H, Kessen W (1964) Uncertainty, structure, and preference. Psychological Monogr: Gen Appl 78(9):1
Murray N, Marchesotti L, Perronnin F (2012) Ava: a large-scale database for aesthetic visual analysis Computer vision and pattern recognition (CVPR), 2012 IEEE conference on, pp 2408–2415
Nadal M, Besch MN, Kelley CS, Mains K, Marty G (2008) The influence of visual complexity on aesthetic preference: An explanation of diverging results. Proceedings of IAEA08 pp 137–141
Nadal M, Munar E, Marty G, Cela-Conde CJ (2010) Visual complexity and beauty appreciation: Explaining the divergence of results. Empir Stud Arts 28 (2):173–191
Nasar JL (2002) What design for a presidential library? Complexity, typicality, order, and historical significance. Empir Stud Arts 20(1):83–99
Nishiyama M, Okabe T, Sato I, Sato Y (2011) Aesthetic quality classification of photographs based on color harmony Computer vision and pattern recognition (CVPR), 2011 IEEE conference on, pp 33–40
O’Donovan P, Agarwala A, Hertzmann A (2011) Color compatibility from large datasets ACM Transactions on graphics (TOG), vol 30-4, p 63. ACM
Oliva A, Mack ML, Shrestha M, Peeper A (2004) Identifying the perceptual dimensions of visual complexity of scenes Proc. of the 26th annual meeting of the cogn. Sci. soc, pp 101–106
Palumbo L, Bertamini M (2016) The curvature effect a comparison between preference tasks. Empir Stud Arts 34(1):35–52
Redies C, Amirshahi SA, Koch M, Denzler J (2012) Phog-derived aesthetic measures applied to color photographs of artworks, natural scenes and objects Computer vision–ECCV 2012. Workshops and demonstrations. Springer, pp 522–531
Rigau J, Feixas M, Sbert M (2008) Informational aesthetics measures. IEEE Comput Graph Appl 28(2):24–34
Romero J, Machado P, Carballal A, Santos A (2012) Using complexity estimates in aesthetic image classification. J Math Arts 6(2-3):125–136
Saklofske D (1975) Visual aesthetic complexity, attractiveness and diversive exploration. Percept Mot Skills 41(3):813–814
Schenkman BN, Jönsson FU (2000) Aesthetics and preferences of web pages. Behav Inform Technol 19(5):367–377
Simond F, Arvanitopoulos Darginis N, Süsstrunk S (2015) Image aesthetics depends on context IEEE Proceedings of the international conference on image processing, EPFL-CONF-212967
Su HH, Chen TW, Kao CC, Hsu WH, Chien SY (2011) Scenic photo quality assessment with bag of aesthetics-preserving features Proceedings of the 19th ACM international conference on multimedia, pp 1213–1216
Sun L, Yamasaki T, Aizawa K (2014) Relationship between visual complexity and aesthetics: Application to beauty prediction of photos Computer vision-ECCV 2014 workshops. Springer, pp 20–34
Tian X, Dong Z, Yang K, Mei T (2015) Query-dependent aesthetic model with deep learning for photo quality assessment. IEEE Trans Multimedia 17 (11):2035–2048
Tuch AN, Presslaber EE, StöCklin M, Opwis K, Bargas-Avila JA (2012) The role of visual complexity and prototypicality regarding first impression of websites: Working towards understanding aesthetic judgments. Int J Hum Comput Stud 70(11):794–811
Van der Helm PA (2000) Simplicity versus likelihood in visual perception: from surprisals to precisals. Psychol Bull 126(5):770
Vu CT, Phan TD, Chandler DM (2012) A spectral and spatial measure of local perceived sharpness in natural images. IEEE Trans Image Process 21(3):934–945
Wakefield KL, Baker J (1998) Excitement at the mall: determinants and effects on shopping response. J Retail 74(4):515–539
Wang L, He DC (1990) Texture classification using texture spectrum. Pattern Recogn 23(8):905–910
Xu Y, Ratcliff J, Scovell J, Speiginer G, Azuma R (2015) Real-time guidance camera interface to enhance photo aesthetic quality Proceedings of the 33rd annual ACM conference on human factors in computing systems, pp 1183–1186
Zhang L, Gao Y, Zimmermann R, Tian Q, Li X (2014) Fusion of multichannel local and global structural cues for photo aesthetics evaluation. IEEE Trans Image Process 23(3):1419–1429
Zhang Q, Shen X, Xu L, Jia J (2014) Rolling guidance filter Computer vision–ECCV 2014. Springer, pp 815–830
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix
ᅟ
A Parameter list
In the extraction of visual complexity features, we set parameters to constant values. We used arithmetic sequence or geometric sequence when setting the parameters to cover the parameter space as much as possible. Further feature selection is helpful to identify the optimal parameter settings. We list the value of these parameters in the following table (Table 6).
B Influence of classifier parameters
3.1 B.1 Subset with consented aesthetic scores
The aesthetics prediction task is implemented by a two-classes-classification method, in which an image is considered as of high quality if its public aesthetics rate is over a certain threshold. In such situation, it is necessary to check the variation among users’ aesthetics judgements to see how reliably the public aesthetics rate could reflect general users’ judgements. We annotate the rates for a certain photograph as \(R^{p_{i}}=\left \{r_{1},r{2},...r_{n} \right \}\), where p i refers to the ith sample in the dataset. We first assume that the distribution of aesthetics rates towards each photograph, follows a normal distribution. And we calculate the standard deviation of the aesthetics rates for each photograph. We consider the accuracy of human labelling as the ratio of the number of photographs, towards which most people agree with the aesthetics quality, in the dataset.
We set the consensus range as \(R_{avg}^{p_{i}}\pm \sigma ^{p_{i}} \), where \(R_{avg}^{p_{i}} \) is the public aesthetics rate, and \( \sigma ^{p_{i} } \) is the standard deviation for the sample. The two standard deviation gap leads to absolute majority (68%) of aesthetics quality judgements. If we divide the samples into two balanced classes, which means we take the average of the public rates for all the photographs in the dataset as the threshold. The threshold is defined as \(thres=avg(R_{avg}^{P})\), where P = p i|i ∈ (0, 255330) is the whole dataset. Photographs with the lower bondary of consensus range \(R_{avg}^{p_{i}}- \sigma ^{p_{i}}\) larger than the threshold, and the photographs with the higher boundary less than the threshold are considered to have converged aesthetics judgements.
In this way we find out that there are only 11198 photographs towards which absolute majority of the participants give a consensus aesthetics judgement, and the gap between the high and low quality of their public aesthetics rates is 1.91, with 6.47 for the lowest rate for high quality and 4.56 for the highest rate for low quality.
3.2 B.2 Aesthetic quality estimation with different classifier
We compare the aesthetic quality estimation accuracy obtained through different classifier with various parameter. As shown in Table 7, SVM with RBF kernel and the classification marginal set as 1000 has the best performance. Adjusting the estimator number of the AdaBoost classifier (base estimator as Decision Tree classifier) does not influence the accuracy very much and the performance is between that of linear SVM and RBF kernel SVM. Using AdaBoost classification, VCPC has the best performance when the number of estimator is set as 400, while VCHA achieve the best when the parameter is set as 800. Generally VCHA features have 1 2 % better accuracy than VCPC features.
Rights and permissions
About this article
Cite this article
Sun, L., Yamasaki, T. & Aizawa, K. Photo aesthetic quality estimation using visual complexity features. Multimed Tools Appl 77, 5189–5213 (2018). https://doi.org/10.1007/s11042-017-4424-4
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11042-017-4424-4