Affine-invariant curve normalization for object shape representation, classification, and retrieval

A novel method for two-dimensional curve normalization with respect to affi ne transformations is presented in this paper, which allows an affine-invar iant curve representation to be obtained without any actual loss of informa tion on the original curve. It can be applied as a preprocessing step to an y shape representation, classification, recognition, or retrieval technique , since it effectively decouples the problem of affine-invariant descriptio n from feature extraction and pattern matching. Curves estimated from objec t contours are first modeled by cubic B-splines and then normalized in seve ral steps in order to eliminate translation, scaling, skew, starting point, rotation, and reflection transformations, based on a combination of curve features including moments and Fourier descriptors.