A study of the impact of image noise on well-known range image curvature determination methods is presented here. The study considers 12 methods, and each is analyzed based on its performance at varying levels of input noise. The performance analyses consider quality factors of (1)
absolute error, (2) correlation with correct, expected curvature values, and (3) signal-tonoise ratio (SNR). Curvature-based renderings are also presented for some data to provide basic visualizations of the impact of noise on one curvature-based task. The work can benefit tasks using range
data (e.g., from Kinect or commercial-grade sensors).