Abstract: Two improvements are discussed with regard to an important type of feature recognition algorithm, which is Falcidieno's algorithm. Firstly, when extracting the explicit features, the original algorithm cannot handle the case when protrusion features and depression features are adjacent. But the improved algorithm can do. It deals with the mixed loop, which is composed of convex edges and concave edges, and then classifies the features into protrusion ones and depression ones. Secondly, a property to extract the implicit protrusion features is discovered. And the method to treat with the implicit features is modified. Hence more kinds of implicit protrusion features can be recognized. These improvements extend remarkably the application range of the algorithm.