We present algorithms for predicting the usefulness and naturalness of color reproductions of natural scenes. The algorithms are based on a computational model of the stages that lead to an observer's impression of the usefulness and naturalness of an image. These stages are (1) the perception, or internal quantification, of color; (2) the construction of a memory standard for an object's color based on its color as observed in the past; and (3) matching of observed object colors with memory standards. In the first of the above stages, the internal quantification of color, the concept of (partially) flexible metrics plays a central role. To test the usefulness algorithm, it was used to predict the discrimination of detail in black and white images of which the contrast was manipulated by applying an s-shaped transform on CIE 1976 lightness L^*. The naturalness algorithm was tested by using it to predict the naturalness of the grass, skin, or sky areas of images of which the color was manipulated by shifting CIE 1976 hue angle h_uv and scaling CIE 1976 saturation s_uv of the grass, skin, or sky areas of the images. The predictions produced in these tests correspond quite well to experimentally obtained judgments of human subjects.