Recently, the importance of information enhancement methods in digital image processing has increased significantly. A large amount of research has been focused on information retrieval and image understanding. Typical examples are searching for similar objects/images in large databases and understanding the objects in images. The main point of these tasks is to extract the most characteristic features of the objects in the images, like edges, corners, characteristic textures, etc. Another very important aspect can be the separation of the "significant" and "unimportant" parts of these features, i.e. the enhancement of those features which carry primary information and to filter out the part which represents information of minor importance. By this, the complexity of the searching and/or interpreting algorithms can be decreased while the performance increased. This paper describes a new edge processing method which is able to extract the "primary", i.e. those edges which can advantageously be used in sketch based image retrieval algorithms.