The work aims to explore the multi-dimensional user perception preference of images in view of the problem that there is a design generation gap between the image expression of traditional products and the user perception, so that the design images can be effectively conveyed. Firstly, the research target was the user perception preference of high-end product images, and a research method of user perception preference combining eye tracking technology and conjoint analysis method was proposed. Secondly, eye tracking technology was used to collect the eye movement state and data when users watched sample pictures. Finally, the conjoint analysis method was used to analyze the important attributes of high-end product images, and the utility value and importance of each attribute were obtained, so as to realize the quantitative analysis of user preferences for high-end product images. At last, the optimal attribute combination scheme that met the user perception preference for high-end product images was obtained. Establishing the research method of correlation between user perception and image attributes of high-end products can help designers optimize product design from the perspective of user perception preferences and improve the scientificness and effectiveness of design.