Non Stationary Wavelets: A New Approach To Best Basis Selection
Khaled Melkemi · 2009
In image and signal processing, the algorithms of compression depend on wavelet bases and their approximations properties. It is a particular and family which gives a sparse representation of piece-wise smooth functions (image or signal). By suitable we mean that the representation allows a simple identification of function's information (e.g. regularity). In this work, we propose a new approach to select the representation for a given function. First, we describe the construction of a wide family of non stationary wavelet bases with specific properties. Then, we give a criterion (and the proof) of best basis selection. Finally, we present some experimental results proving the effectiveness of our approach.