A geometrical a priori for capturing the regularity of images

We first briefly recall how to model occlusion and scaling in natural images through the use of a stochastic model, the scaling dead leaves model. Then we give a statistical estimator for its scaling parameters, which are related to the regularity of images. Last we show how this model can be used as an a priori for image denoising, in the framework of wavelet coefficients thresholding.