Feature detection is important in computer vision. Many methods have been proposed that attempt to detect features in images automatically. The method proposed by Per-Erik Danielsson is investigated in this thesis. It uses second derivative information to extract local shape and orientation from a grey-scale image. The algorithm returns a parameter that classi es the types of shape including blobs, ridges, lines, and saddles present in the image. One of the implementation issues addressed is the algorithm's many-to-one shape mapping, where di erent image features are mapped to the same shape parameter. The thesis explains how to determine correct results from the shape parameter. Another issue that is addressed is the choice of scale of the lters used. The proposed solution applies several lters at di erent sizes, then combines the responses using the best scale for each pixel. The thesis also shows how to visualise the information returned by the algorithm. This is done by colouring an image according to the shape parameter. Canonical shapes, blobs, ridges, lines, and saddles, being represented by red, green and blue. Finally, the method is applied to mammograph images to highlight important features of each image.