This paper presents a neural network (NN) approach to determining the best matching colors for a given product image in product design. With a wide variety of forms and colors on the market, mobile phones are used in an experimental study to illustrate the approach. 33 representative mobile phones and 50 commonly used colors are used as experimental samples to examine how product color affects a product image perceived by consumers. Four NN models are built with different hidden neurons to examine how a particular combination of mobile phone form and color matches the simple-complex image. The performance evaluation result shows that the number of hidden neurons has no significant effect on the predictive ability of the NN models. The NN models can be used to construct a product color database for supporting design decisions on product colors for matching a desirable product image.