Color image analysis is still limited by the very significant amount of data. Most real-world images are, of course, not monochrome, but full color. Three-dimensional imaging utilizes large data sets, demanding computer storage and speedy algorithms of critical importance. With this aim in view, without taking into account both spatial and color information, a reducing step cannot be efficient. Thus, a multiresolution process seems to be well adapted, allowing simpler and faster computations. We present here a multiresolution tool, the Gaussian pyramid, first introduced for gray-scale images. We discuss the way to construct the color pyramid in a gamma-corrected RGB space, where the color mixing is additive.