Let P be a set of points in a d-dimensional space and let 0< e < 1 a given parameter. We will show that is not possible in general to find a subset Q' of P of size independent of |P| such that: |MST(Q')| >= (1 - e)|MST(P)|. Where |MST(P)| denotes the weight of the euclidean minimum spanning tree of P. On the positive side, we will show the existence of a subset Q of P such that its size is independent of the size of P and such that a spanning tree ST(Q) and MST(P) are close in the Hausdorff metric, that is, they look essentially equal. This might have applications in Image Comparison, Pattern Recognition and actually gives a way to compress MST(P) as the space required to store ST(Q) is independent of |P|. Further, we also give possible approaches to compute such a spanning tree ST(Q) and therefore Q under any given fixed metric Lp, 1 <= p <= infty, and any fixed dimension d.