Abstract: This study explores the mechanism of the influence of the traffic noise on the mental health of the residents in the residential area along the expressway, and focuses on the mediating role of the landscape perception. By employing virtual reality (VR)-constructed landscape scenarios with quantitatively controlled audiovisual parameters, collecting data through 5-point Likert scales for audiovisual perception and standardized mental health questionnaires, the experiment examined the relationships between environmental features, perceptual responses, and mental health under 67 dB traffic noise exposure. Using a randomized controlled trial with a 2 × 2 factorial design (high/low Green View Index (GVI) × presence/absence of natural sounds), the subjective feedback results of audiovisual environmental characteristics were recorded, while participants’ electrodermal activity were simultaneously monitored. Results revealed that high-GVI landscapes with natural sounds significantly improved mental health outcomes (p < 0.05). While no significant interaction was observed between visual and auditory environmental features (p > 0.1), a robust cross-modal interaction effect emerged between visual and auditory perception (p < 0.01), particularly enhancing positive affect. The “landscape characteristic–audiovisual perception–mental health benefits” structural equation model demonstrated that subjective perception fully mediated the pathway from environmental characteristics to mental health (T = 3.479, p < 0.01), confirming a complete mediation effect. These findings indicate that residents’ mental health in noise-exposed environments is modulated by audiovisual perceptual quality, suggesting that optimized multisensory environmental design can enhance perceptual experiences and consequently improve psychological well-being.