Leaf margin feature is one of the important bases in plant species identification and plant classification. A scheme of computer aided feature extraction of leaf margin has been designed based on descriptions of leaf margin in botany. Firstly, the leaf teeth are detected using an improved SUSAN algorithm. Then three feature parameters of leaf teeth are computed, which are the numbers, sharpness and inclination of the leaf tooth. The scheme was tested on a few types of leaf margin. The result showed that the improved SUSAN algorithm was a useful tool in leaf tooth detection, the three parameters reflected the shape difference among different types of leaf margin and the scheme could be applied in quantitative classification of leaf margin.