Comparison of built-up indices for urban features extraction

A comparative evaluation of built-up indices applied to high-resolution multispectral imagery from the WorldView-2 satellite is presented. Four metropolitan areas are analyzed: Athens, Bucharest, Rome, and Turin. Each index is tested on different land covers, such as vegetation, water, bare soil, and built-up. A binarization strategy based on cumulative histograms of ground-truth samples is implemented. Performance was assessed using the Spectral Discrimination Index (SDI). The results indicate that NBEI outperformed the other indices in separating impervious surfaces from spectrally similar bare soil. The findings confirm the potential of index-based approaches for built-up extraction in high-resolution imagery, while also highlighting their limitations in heterogeneous urban environments.