ABSTRACT
This paper describes the development of a computational model for the estimation of a comfortable amount of information in motion graphic videos; image structure analysis is used in the development. A time series of frame differences (TSFD) is used as a computable index of the amount of motion in order to investigate the correlation between the spatiotemporal feature of motion and subjective impressions, including “activity,” “complexity,” and “comfort.” By performing correlation analyses between impression scores based on the results of a subjective impression evaluation and the power ratios of the time series of frame differences, it is suggested that each impression has a correlation with a particular band of spectrum of TSFD. These bands correspond to human information processing characteristics. The retention time of VIS and Visual STM and the property of pattern chunking have a strong relation with the impression evaluation of a structural feature of motion graphics.
