Highlights are due to specular reflection and cause the lustrous or mirrorlike appearance of many material surfaces. We investigated in detail the structure of highlight patterns that are due to material surface roughness. We interpret results in terms of a simple model of a random Gaussian surface. The model’s prediction corresponds with the microscopic measurement within a factor of 2. The method allows one to rank generally the roughness of the surfaces of the fruit samples by purely optical means. This simple procedure for estimating surface roughness from images has implications for visual perception and graphic rendering.