Abstract: In recent years,computational aesthetics has become an active research field.Mobile device applications(APP)have covered every aspect of life and become an indispensable part of people’s daily life.Unfortunately,few attempts have been made toward the computational evaluation of APP icon aesthetics.A framework for computational aesthetic evaluation of APP icons is proposed.A set of feature variables is defined to evaluate some aspects of APP icons,such as colorfulness,harmony and contrast.The human subjective ratings of 60 volunteers on the colorfulness,harmony,contrast and aesthetics of APP icons are also collected,and the statistical linear regression models are constructed.The results show that proposed regression models can explain 84%-91%of human judgments on colorfulness,harmony,contrast and aesthetics of APP icons,and the evaluation scores of proposed models are highly correlated with manual evaluation scores.