A novel multifractal image denoising approach based on singularity structure analysis is proposed to overcome the difficulty in estimating the multifractal spectrum of discrete data for multifractal analysis. By defining the partition function for moments of the wavelet coefficients

a new estimator for multifractal spectrum is given. There is no assumption for the type of noise or global smoothness of original data

and the image is characterized by its multifractal spectrum. The Hausdorff exponent of the image is processed by defining a new transformation operation

so that most of the points in the denoised image lie in the smooth regions while the relative strength of the singularities is preserved. Experimental results show that the method proposed leads to a smooth image and preserves the texture information of original images very well.