The shape is one of the most popular representations of an object. Algorithms of extracting object’s shape are well known and widely used. Furthermore, there are many descriptors based on information about silhouette. However, there are some problems to overcome, for example noise, occlusion, distortion caused by affine transformations. This paper presents experimental results of applying the UNL-Fourier Descriptors to contour objects. The approach utilizes the transformation from Cartesian to polar coordinates system (it is close to so-called “signatures”) ith normalization, and subsequently, Fourier transform, which can be used to feature reduction. It is worth to note, that the UNL-Fourier Descriptor belongs to larger class of affine invariants.