People prefer a visually appealing interface. Aesthetics, as a significant factor of user subjective satisfaction, further influence the usability as well. Recently, HCI researchers focus on evaluating UI aesthetics automatically by applying computational metrics for this evaluation. When adapting the metrics for image quality and web page complexity evaluation to mobile interface, the explanation ability of this model is greatly declined.
To explore the possible breakdowns and the potential solutions, this paper analysed 21 pixel based computational metrics with 102 mobile UIs from three aspects: 1. descriptive statistics to understand the data mathematical features; 2. correlation and factor analysis to learn the relationship between metrics and 3. the computing performance. The results reveal the possible influencing factors of metrics, such as the metrics being highly skewed or slightly diverse because of the photograph, algorithm parameters, image resolution, color reduction, and even the brightness of the screen. Finally, we suggest the directions of possible solutions about data preprocessing and metric improvement. To have a more accurate model, the various standards could be applied to bitmap based interfaces and vector graphics based interfaces.
