Toward Model Building for Visual Aesthetic Perception

Table 2

Comparative analysis of the six reviewed models.

Model name	Objective of the model	Input stimulus	Are there some structural parameters that cannot be quantified for the development of mathematical models?
Neuropsychological model [6]	The neuropsychological model is derived from visual cognitive neuroscience and aims to propose a general framework for the neural underpinnings of visual aesthetics and to explain how visual stimuli are mapped to emotions.	Visual information	Nearly all the visual computation stages involved in this model can be mathematically simulated. There are some models and algorithms for visual feature extraction. This model is easily acceptable for model building.
Information-processing model [7, 8]	By positing several distinctive processing stages among cognitive and emotional processes, information-processing models have been proposed for empirical research on the perception of art and aesthetic experiences.	Modern and contemporary visual art, everyday objects, design objects, dance and body perception, music, and food	Some structural parameters, such as the time course of early processes, the relevance of prior experience, the complexity and relevance of emotional processes, and life-relevant experiences, are difficult to quantify.
Mirror model of art [9]	The mirror model of art focuses on the interface between the creation and appreciation of art. The model is based on the assumption that the fundamental nature of art-making and art-viewing is related.	Visual arts	The knowledge and background experiences of the artists and perceivers when they are creating and viewing art are difficult to describe with computational variables.
Quartet model of human emotion [10]	The quartet model of human emotion presents a neurobiological theory of emotions and proposes four core emotional systems. Aesthetic perception is also covered in the quartet model.	Everything related to emotion, including language	The quartet model of human emotion focuses on the interactions and connections among the four core systems. How to model the interactions among these four systems using mathematical functions is a great challenge.
Unifying model of visual aesthetic experience [11]	The model aims to explain factors involved in aesthetic experiences in response to visual artworks.	Visual stimuli	The forward flow of information in different stages and the encoding of information in the nervous system can be expressed by mathematical functions. However, the beauty-responsive mechanism and the personal cultural filter are difficult to describe using mathematical models.
Hierarchical “feed-forward” model [12–14]	The hierarchical feed-forward model was developed using multiple linear regression to investigate the relationship between human aesthetic texture perception and computational low-level texture features. These white-box models can be interpreted in terms of both structure and interactions between aesthetic properties and texture features according to feature weights.	Visual stimuli	The models can be expressed with mathematical functions. However, the models do not accurately reflect the real functions of the neural systems involved in the aesthetic perception of visual art.