A visibility-aware framework for façade-level building height estimation using multi-view street-view imagery and building footprints

Building height is a key indicator of urban vertical structure, yet most existing methods still estimate one representative height for each building. Although adequate for many applications, this representation cannot localize height variation along the building’s façade edges in stepped or other multi-height structures, limiting its use in street-scale analyses that depend on local building geometry, such as shadow, noise, and visibility modeling. To represent such localized variation, this study proposes a multi-view computational framework to estimate façade-level building heights, based on panoramic street-view imagery (SVI) and building-footprint GIS data. The framework combines visibility-aware SVI-façade matching, projection-guided pixel-to-façade modeling, and view-quality-aware multi-view fusion using pointwise weighted median aggregation to capture façade-level height variations. Evaluations in Rotterdam and Amsterdam showed that the proposed framework outperformed both evaluated baselines while additionally preserving façade-level height variation. On the common detected building sets, it achieved mean absolute errors (MAEs) of 1.13 m and 3.50 m, respectively, reducing errors by 22.1% and 27.4% relative to Pano2Geo and by 74.6% and 79.4% relative to SIHE. It also achieved the highest proportions of estimates with absolute errors below 1, 2, and 3 m in both sets. At the façade level, MAEs were 1.09 m in Rotterdam and 3.68 m in Amsterdam against the official LoD 1.3 reference. The proposed method retained the lowest building-level MAE across all occlusion groups in both areas, although high-rise buildings with slanted roofs remained challenging. These results indicate that the proposed framework can recover localized street-facing height variation from multi-view SVI and building footprints while retaining accurate building-level estimates after aggregation. The resulting street-facing height information provides a geometric basis for detailed urban digital twins and could support street-scale analyses of shadow and solar exposure, noise propagation, and visibility or street enclosure.