In order to improve the evaluation approach of only a single visual kansei perception, this study proposes an evaluation model of product compound kansei image. The implementation procedure first identifies product samples and image words, and conducts a kansei measurement questionnaire. The obtained data were analyzed for the important weightings of kansei words using entropy method. The approach to determine weightings by the divergent degree of performance scores can enhance rationale and objectivity, and thus reduce the shortcoming of subjective weighting setting. The TOPSIS approach was then used to rate the overall performance under a compound kansei image. According to the relative c10seness degree, the ranking order can be achieved. The results can assist designers to establish benchmark of product form style and develop conceptual design with varieties.