Texture characterization, which measures the fluctuations of image amplitude regularity in space, now often uses fractal and multifractal analysis. The present contribution deals with a multifractal approach in ultrasound skin images to characterize melanoma. In this paper, we propose to build new hierarchical multiresolutions quantities: Maximum coefficients of the Discrete Wavelet Transform for 2D Multifractal analysis. The performance of the proposed texture descriptors was evaluated for 2D synthetic processes and biomedical texture classification.