Moment Invariants are one of the techniques of feature extraction frequently used for shape recognition algorithms. A moment is a projection of function into polynomial basis and Moment Invariant is a moment function retaining invariance to particular class of degradation (e.g. affine transformation, convolution with symmetric kernel, etc.). Moment Invariants are thoroughly studied objects, but mainly in the continuous representation. The transition to discrete domain have many effects, some of which are not covered in literature, but potentially having profound impact in pattern recognition applications. Typical example of such an effect is dependence of features focused by this project. Several techniques of moment invariant creation exist often generating overcomplete set of invariants. Dependencies in these sets are commonly in a form of complicated polynomials, furthermore they can contain dependencies of higher orders. These theoretical dependencies are valid in the continuous domain but it is well known that in discrete cases are often invalidated by discretization. It is likely, that it would be feasible to begin classification with such an overcomplete set and adaptively find the pseudo-independent set of invariants by the means of feature selection techniques.