The up-close experience of perennial landscapes has been shown to enhance residents’ perception of naturalness in the context of increasing small-scale vegetation landscapes. This study explored how formal aesthetic characteristics were related to landscape perception and whether landscape preference correlated with eye movements. We created a series of photomontages showing perennial combinations that contained different plant forms, degrees of species richness, and plant arrangements and recorded 73 participants’ eye movements during 10 s of free viewing in Experiment A and task-oriented viewing in Experiment B and ratings of landscape preference collected through rating scales. We found that the effects of plant form and species richness were significant for gaze behavior, while arrangement showed no significant effect. We also found that landscape preference was positively correlated with fixation count but negatively correlated with mean fixation duration and total fixation duration. Additionally, women had more but shorter fixations than men while viewing these photomontages, and the difference in aesthetic preferences between men and women was not significant. Concerning the different professional background groups, no significant gaze behavior difference between professionals and nonprofessionals was detected, but compared with professionals, nonprofessionals tended to give strongly higher preference ratings. The outcomes shed light on the influence of formal aesthetic characteristics on gaze behavior and advanced the application of eye-tracking technology in perennial landscape studies. Our findings also confirmed the efficiency of vegetation landscapes designed based on public preferences for providing restoration from stress or fatigue.