Urban green spaces play a crucial role in providing social services and enhancing residents’ mental health. It is essential for sustainable urban planning to explore the relationship between urban green spaces and human perceptions, particularly their visual comfort. However, most current research has analyzed green spaces using two-dimensional indicators (remote sensing), which often overlook human visual perceptions. This study combined two-dimensional and three-dimensional methods to evaluate urban green spaces. Additionally, the study employed machine learning to quantify residents’ visual comfort in green-space environments and explored the relationship between green spaces and human visual perceptions. The results indicated that Kitakyushu exhibited a moderate FCV and an extremely low Green View Index (GVI). Yahatanishi-ku was characterized as having the highest visual comfort. Tobata-ku demonstrated the lowest visual comfort. Natural, GVI, openness, enclosure, vegetation diversity, landscape diversity, and NDBI were positively correlated with visual comfort. FCV and ENVI were negatively correlated with visual comfort. Vegetation diversity had the most impact on improving visual comfort. By integrating remote sensing and street-view data, this study introduces a methodology to ensure a more holistic assessment of green spaces. Urban planners could use it to better identify areas with insufficient green space or areas that require improvement in terms of green-space quality. Meanwhile, it could be helpful in providing valuable input for formulating more effective green-space policies and improving overall urban environmental quality. The study provides a scientific foundation for urban planners to improve the planning and construction of healthy and sustainable cities.