Color images are usually observed as integrated objects, or called color gestalt. That is why color effect of image is not only affected by the color components themselves, but also determined remarkably by the image constructive frame, which holds the colors. Although, previous researchers have already explored this issue, the conclusions are usually descriptive principles that one should keep in mind for color designing. Precise quantification for image composition of color gestalt is still an ignored field in color research domain. As the use of computer-aided design (CAD) increases in the color design field, more quantifiable measures for constructive feature factors that affect color gestalt are needed. In this article, we introduce a three-dimensional model of image composition that can both quantify the effect of color gestalt and rebuild image composition to acquire certain color effect for color combination. This method pioneers a new perspective of appreciating the influence of image construction on color combination effect, and a new way of researching color by affective computing or Kansei engineering technology. To test the feasibility of the scale model by a usage study, an interactive toolkit was also built. Practice has confirmed that this method is a technical meaning in application technically meaningful approach, and will spur further research on color composition. (c) 2015 Wiley Periodicals, Inc. Col Res Appl, 41, 165-174, 2016