Any milk drinker knows that there are three basic varieties. There is whole milk, favored by some for its rich, almost creamy, taste. Then there is regular milk, its diluted half brother. It is noticeably thinned, with maybe just a trace of cream. And, finally, there is skim milk, the watery and weak stepchild familiar to those who are weight- and health-conscious. Besides the taste, the appearance of milk also differs depending on its content. But until now, these subtleties in texture could not be rendered easily by computer graphics. In most images, milk was just milk, a flat white liquid in a glass.These rendered images show the components in milk and in mixed concentrations of milk. Starting at the left, the glasses contain water, water and vitamin B2, water and protein, water and fat, skim milk, regular milk and whole milk. Courtesy of the University of California, San Diego.Researchers at the University of California, San Diego, have created a computer graphics model that changes that. It generates highly realistic images of milk varieties by factoring in its fat and protein content. Change the parameters, and the texture of the image changes accordingly. The model achieves this by determining how light interacts with the specified ratio of fats and proteins. The system also can take a digital picture of milk and, working backward, determine its fat and protein content. The same technique works with ocean water. Specify the algae content and the types of minerals in the water, and the model will generate images that range from an inviting Mediterranean blue to an unappealing brackish green. Indeed, the model works with a host of participating mediums, where some of the light that hits a material is absorbed and not reflected. Computer professor Henrik Wann Jensen said that his technique has a number of potential applications, including the identification of spoiled or contaminated food. One company has expressed interest in using the model to check the freshness of milk and ice cream. Beyond milk and water, Jensen is working to extend the model’s applications to generating images of human skin. It would be able to predict the appearance of skin based on a detailed description of the dermal structures.