Density estimation arises in a wide range of vision problems and methods which can deal with high dimensional image features are of great importance. While in principle a non- parametric distribution can be estimated for the full featu re distribution using Parzen win- dows technique, the amount of data to make these estimates accurate is usually either unattainable or unmanageable. Consequently, most modelers resort to parametric models such as mixtures of Gaussians (or other more complicated parametric forms) or make in- dependence assumptions about the features. Such assumptions could be detrimental to the performance of vision systems since realistically, image f eatures have neither a simple para- metric form, nor are they independent. In this paper, we revive non-parametric models for image feature distributions by finding the best tree-structured graphical model (using the Chow-Liu algorithm) for our data, and estimating non-parametric distributions over the one- and two-node marginals necessary to define the graph. This procedure has the appealing property t hat, if the tree-structured model represents the true conditional independence relations fo r the features, then our estimated joint distribution converges rapidly to the true distribut ion of the data. Even when this is not true, it converges to the best possible tree-structured model for the original distribution. We illustrate the effectiveness of this technique on simula ted data and a real-world plankton classification problem.