This paper presents a discrete structure, named adaptive tangential cover (ATC), for studying 3D noisy digital curves. The structure relies mainly on the primitive of blurred segment of width ν and on the local noise estimator of meaningful thickness. More precisely, ATC is composed of maximal blurred segments of different widths deduced from the local noise values estimated at each point of the curve. Two applications of ATC for geometric estimators of 3D noisy digital curves are also presented in the paper. The experimental results demonstrate the efficiency of ATC for analyzing 3D irregular noisy curves.