The problem of retrieving spatial information of sea surface heights from aerial images is considered. In this paper, some analytical and numerical results that relate the autocorrelation of the surface heights and those of the sun glint patterns are derived. Examples of these results are presented showing the nonlinear technique which can be applied to obtain the power spectrum of a real sea surface using aerial photographs. As a first step, two kinds of roughness spectra are presented: two roughness spectra which describe single-scale surfaces, and the Pierson-Moskowitz power spectrum which describe multiscale surfaces for a fully developed sea. Both simulations are presented in one dimension. However, a single-scale bidimensional surface is analyzed considering a Gaussian roughness spectrum. Wave height spectra re obtained from the surface height autocorrelation via a Fourier transform. The results of the model compared with the theory and the data are in quite good agreement. Under favorable conditions, it is possible to invert the relation numerically and estimate the wave height spectrum from the sun glint data. ¿ American Geophysical Union 1993