We present a new method for representation of Multiresolution graphic models. The considered models are curves, images and surfaces. The method is based on a discrete approach to the theory of wavelets. Multiresolution has various applications to computer graphics and pattern recognition, such as image compression and smoothing, graphics model processing, and scanning transformations. We concentrate on the Chaikin"s subdivision method for generating smooth models. This method makes use of an optimal projection of a Multiresolution model onto a space having a lower Multiresolution level. The optimization is made locally and discretely and results in biorthogonal wavelets. We demonstrate the effectiveness of the decomposition and reconstruction processes of our method through various examples.