The purpose of this study was to characterize the natural statistics of surgical images as a first step in exploring the application of methods from vision science to the problem of tissue discrimination in surgery. Field has previously shown that natural images have a simple characteristic spatial structure, with amplitude spectra that decrease with frequency roughly as 1/f. Several subsequent studies have shown that amplitude is generally thought to be proportional to 1/fα where α has been found to be within a fairly narrow range (0.7 -1.5) for natural scenes. Four sets of surgical images were acquired for the current study and their spatial frequency content analyzed. The value of α for these images was found to be 1.59. The same measure for a set of 60 natural images from the Brodatz book of texture gave a value of 1.15. Consequently, we developed examples of image enhancement using a modified histogram equalization technique and a non-linear normalization transform. We conclude that surgical images lack high spatial frequency content, but it is feasible to enhance target structures.