This chapter considers the early stages of visual processing, the nature of the initial representations built from a retinal image, and the algorithms that produce them. In Chapter 1, we explained how light provides information about the visual world, through the relationship between the spatial structure of the optic array and surrounding surfaces and objects. A promising starting point for processing the retinal image, which is a projection of the optic array, is therefore to create a representation of it that makes its spatial structure explicit, specifying where the most significant changes occur in the intensity and spectral composition of light. Notice that we are making a simplifying assumption here, treating the retinal image as static and ignoring change in its structure over time caused by movement of the perceiver or of objects. In Chapter 8, we will return to the question of how motion in a retinal image can be processed and represented.