This paper presents a novel technique for simplifying a triangulated surface at different levels of resolution. While most existing algorithms, based on iterative vertex decimation, employ the distance for error metric, the proposed algorithm utilizes an edge criterion for removing a vertex. An interior angle of a vertex is defined as the maximum value of all possible angles formed by combinations of edges connected to a vertex. Since the surface curvature examined with the interior angle provides more information for decision of vertex removal than the conventional distance measure, the proposed algorithm can approximate the surface with less computation. The height of a triangle, which is formed by the pair of edges, is also used for an additional constraint. The computational complexity can thus be greatly alleviated to logarithmic scale from the exponential scale required for the conventional algorithms, while yielding the comparable error level.