We introduce a new, to our knowledge, method using wavelets and probability theory for the evaluation of speckle interference patterns for quantitative out-of-plane deformation measurements of rough surfaces of nontransparent solids. The experiment uses a conventional Twyman-Green interferometer setup. The speckle interference patterns are obtained by the common method of subtraction of images taken before and after a surface deformation. The data are processed by a wavelet transformation, which analyzes the image structures on different length scales. Thus it is possible to separate the interference fringes from the noise. From the locations of the interference fringes, the deformation of the surface can be reconstructed by means of probability theory.