Automatically evaluating the aesthetic qualities of a photograph is a current challenge for artificial intelligence technologies, yet it is also an opportunity to open up new economic and social possibilities.
Aesthetics in Digital Photography presents theories developed over the last 25 centuries by philosophers and art critics, who have sometimes been governed by the objectivity of perception, and other times, of course, by the subjectivity of human judgement. It explores the advances that have been made in neuro-aesthetics and their current limitations.
In the field of photography, this book puts aesthetic hypotheses up against experimental verification, and then critically examines attempts to "scientifically" measure this beauty. Special attention is paid to artificial intelligence techniques, taking advantage of machine learning methods and large databases.
Table of Contents
Introduction: Image and Gaze xi
Chapter 1 The Legacy of Philosophers 1
1.1. The objectivist approach 3
1.1.1. The source: ancient Greece 3
1.1.2. After Greece 5
1.1.3. Kant and modern aesthetics 7
1.1.4. Objectivism after Kant: from pseudo-subjectivism to aesthetic realism 9
1.2. The subjectivist approach 13
1.2.1. From classicism to romanticism 14
1.2.2. The moderns 15
1.2.3. The influence of neurobiology 18
1.3. Subjectivism and objectivism: an ongoing debate 19
Chapter 2. Neurobiology or the Arbitrator of Consciousness 25
2.1. fMRI protocols and neuroaesthetics 27
2.2. The fMRI quest for “beauty processes” in the brain 28
2.2.1. The role of the prefrontal cortex 28
2.2.2. The role of the insular cortex 31
2.2.3. The role of the visual areas 33
2.2.4. The role of memory and cognition 35
2.2.5. The role of embodiment 35
2.3. Responses from functional electric encephalography 36
2.4. A global cognitive scheme for aesthetic judgment? 39
2.4.1. J. Petitot’s neurogeometric model 40
2.4.2. A. Chatterjee’s aesthetic emotion model 40
2.4.3. The model by Brown et al. 42
2.4.4. Model proposed by H. Leder 43
2.4.5. The model by C. Redies 45
2.4.6. The emotions model developed by S. Koelsch et al. 47
2.4.7. L.H. Hsu’s model of emotions based on A. Damásio 47
2.4.8. Other models 50
2.5. A critique of neuroaesthetic methods 51
2.5.1. Criticism of neuroaesthetic methods 51
2.5.2. Criticisms of the objectives of neuroaesthetics 52
Chapter 3. What Are the Criteria For a Beautiful Photo? 55
3.1. Before we enter into the fray 56
3.1.1. What reference books do we have? 56
3.1.2. “Beauty of an image” or “quality of an image”? 57
3.1.3. A glossary of aesthetic appraisal 58
3.1.4. Measuring beauty 60
3.2. Composition 63
3.2.1. Complexity versus simplicity 63
3.2.2. Unity 64
3.2.3. A specific case in composition: landscapes 64
3.2.4. Using oculometry to analyze composition 67
3.2.5. Format or aspect ratio 68
3.2.6. The rule of thirds (RoT) 70
3.2.7. The center of the image 72
3.2.8. Other rules for composition 73
3.3. Histograms, spectral properties and textures 76
3.3.1. Histograms and gray levels 76
3.3.2. Focus, spectral density, fractals 78
3.3.3. Textures 80
3.4. Color 82
3.4.1. About the concept of color 82
3.4.2. Preferences related to isolated colors 84
3.4.3. Preferences related to color palettes 86
3.5. What behavioral psychosociology has to say 93
3.5.1. Images of nature 93
3.5.2. The aesthetics of faces 96
3.5.3. The role of the signature, title and context 99
3.5.4. Perception and memory: prototypicality 101
Chapter 4. Algorithmic Approaches to “Calculate” Beauty 103
4.1. First steps: C. Henry 103
4.2. G.D. Birkhoff’s mathematical approach 104
4.3. Those who followed G.D. Birkhoff 106
4.3.1. Beauty according to H.J. Eysenck 106
4.3.2. The Post-War years: the designers, A. Moles and M. Bense 106
4.3.3. A dynamic approach: P. Machado and A. Cardoso 107
4.3.4. Work carried out by J. Rigau, M. Feixas and M. Bert 108
4.4. Algorithmic approach with AI: J. Schmidhuber 110
Chapter 5. The Holy Grail of the Digital World: Artificial Intelligence 113
5.1. Which artificial intelligence? 114
5.1.1. The principles 114
5.1.2. Learning algorithms 115
5.2. Why artificial intelligence in aesthetics? 116
5.3. Expert opinions 118
5.4. The database 120
5.4.1. Generalist databases, used for aesthetic judgments 122
5.4.2. Databases that are specialized for aesthetic photography 126
5.4.3. Databases dedicated to artistic judgment 129
5.4.4. Other image databases that are sometimes used 130
5.4.5. Increasing databases 131
Chapter 6. Primitive-based Classification Methods 133
6.1. Judging aesthetics 136
6.1.1. Multimedia primitives: the ACQUINE system (Datta et al.) 136
6.1.2. Edges and chromatic distance: Ke et al. 137
6.1.3. Photography rules: Luo and Tang and Mavridaki and Mezaris 140
6.1.4. High-level primitives: Dhar et al. 143
6.1.5. Generic descriptors of vision: Marchesotti et al. 144
6.2. Help in composing beautiful photos 148
6.2.1. The library of aesthetic primitives developed by Su et al. 148
6.2.2. The OSCAR system by Yao et al. 148
6.2.3. Embedded systems: Lo et al. and Wang et al. 150
6.3. Some specific research related to the evaluation of aesthetics using primitives 151
6.3.1. Color harmony: Lu et al. 151
6.3.2. Group photography: Wang et al. 153
6.3.3. Social networks and crowdsourcing: Schifanella et al. 153
6.3.4. Looking at comments: San Pedro et al. 154
Chapter 7. Deep Neural Network Systems 155
7.1. DNNs dedicated to aesthetic evaluation 157
7.1.1. High and low resolutions: the RAPID system, Lu et al. 157
7.1.2. The multi-path DMA-Net architecture: Lu et al. 160
7.1.3. Adapting to the size of the image: Mai et al. 160
7.1.4. Finding beauty on the Web: Redi et al. 163
7.1.5. Siamese and GAN networks: Kong et al. and Deng et al. 165
7.1.6. Paying attention to the image construction: A-Lamp 167
7.2. Variants around the basic DNN architecture 170
7.2.1. Comparing photos between themselves: Schwarz et al. 170
7.2.2. Making use of knowledge of the subject: Kao et al. 172
7.2.3. BDN: halfway between classification and DNN 174
7.2.4. Using the distribution of the evaluations 175
7.2.5. Extracting a “dramatic” image from a panorama: the Creatism system 178
7.3. Written appraisals: analyzing them and formulating new ones 179
7.3.1. Photo critique captioning dataset (PCCD) 181
7.3.2. Neural aesthetic image retriever (NAIR) 182
7.3.3. Semantic processing by Ghosal et al. 182
7.3.4. Aesthetic multi attribute network (AMAN) 183
7.4. Measuring subjective beauty 185
7.4.1. Recommendation systems 186
7.4.2. Defining the user’s psychological profile 188
7.4.3. Learning the user’s tastes through tests 191
7.4.4. Multiplying concurrent expertise 194
Chapter 8. A Critical Analysis of Machine Learning Techniques 197
8.1. The popularity of studies on aesthetics 197
8.2. A summary of learning methods 199
8.2.1. Which architecture? Which software? 199
8.2.2. What performances? 200
8.3. Questioning the hypotheses 203
8.4. Specific features of beautiful images detected by a computer 204
8.4.1. Some observations on the photos in the AVA database 205
8.4.2. The scores in the AVA database 207
Conclusion 213
Appendix 1. A Brief Review of Aesthetics 219
Appendix 2. Aesthetics in China 237
Appendix 3. The Aesthetic of Persian Miniatures 251
Appendix 4. Aesthetics in Japan 263
References 271
Index 295
