Forest colors are important elements for public enjoying the scenery. So increasing attention has been acquired on forest color cognition. However researches on relationship between forest colors and public response are still insufficient, which cannot provide sufficient theoretical basis for the regulation of forest landscape in color. Therefore, We seek to examine the relationship between forest color and visual behavior based on eye tracking technology, and further interpret the visual indicators through the value of scenic beauty. This study researched Jiaozi Mountain in China by selecting 29 sampling points, counting up 116 photographs in 4 seasons by a mountainous region. On this basis, Matlab was performed to quantitatively extract color elements, while ArcGIS and Fragstats were applied to extract the spatial index of color patches. A total of 10 indicators were obtained to explain the color characteristics of each forest image. Through both visual behavior experiment and landscape preference evaluation, the results showed that people tend to have different visual behaviors and preference cognition when observing forest colors of different seasonal types. Based on the study of forest landscape color in all seasons, the subjects tend to judge the image in comparison to other seasonal forest landscape color photos to identify it more easily. For a single-season forest colors, diversified color information and abundant visual attention are important factors influencing the correlation between visual behavior, landscape preference, and forest color characteristics. This study aims to further reveal people's perceptions and psychological preference to forest colors, contribute to the establishment of a more quantitative and scientific scenery evaluation system, and provide a scientific basis for forest color planning and design.