A well-known method for generating computer art is based on the paradigm of user-guided evolution, a method whereby "artists" interactively search through populations of images in order to select, and subsequently breed, those images which show aesthetic promise. This can be a time consuming endeavor, lacking well-defined principles for controlling evolution and for obtaining images with desired aesthetic characteristics. One way to improve upon this situation is to develop computational criteria for selecting images that conform to user specified aesthetics. We consider a non-interactive evolutionary method that contributes to this field of computational aesthetics by: (1) color segmenting the digital images that we breed so that their color organization can be used to influence aesthetic decision making, and (2) investigating mathematical models for assigning aesthetic fitness to images based on geometric quantities obtained following color segmentation. We provide examples of the aesthetic imagery we obtain.