Analyzing Urban Visual Environments through Spatial Patterns of Color from Street View Imagery

Abstract. Urban environments are experienced not only through their density, land use, and spatial layout but also through their visual character. Color, for instance, is a prominent feature of façades and streetscapes, affecting the perception, memory, and value of cities. Nevertheless, despite its significance, widespread patterns of urban color have been understudied, and their relationship with urban function has rarely been systematically explored. This study provides a methodology combining street-view imagery, deep learning-based semantic segmentation, and perceptual color indices to evaluate façade colors in Quezon City, Philippines. From 42,942 Google Street View images, façades were segmented with the SegFormer model, and four indices—vibrancy, harmony, tone balance, and entropy—were computed to capture complementary dimensions of visual character. The results show clear functional signatures: commercial areas are marked by high vibrancy and entropy, residential areas show higher harmony, institutional complexes show homogeneous color schemes, and industrial parks show muted tones. Road types also show variation, with trunk and tertiary roads showing visual diversity, while motorways and service roads show homogeneity. Furthermore, hotspot analysis shows contrasts between informal settlements, upscale subdivisions, and industrial corridors. These results demonstrate that color provides valuable information on the interdependence of visual character and urban function. By placing color in the position of a quantifiable dimension of urban form, the study highlights its ability to inform planning and design policies to support livable, unique, and sustainable urban environments.