Method and apparatus for edge detection in images

Abstract

Abstract Conventional methods for determining edges in images operate with linear or non-linear but in any case two-dimensional operators in a filtered image. During this process, however, some edges or edge sections are not detected so that the edges have gaps and due to the filtering, the edge found is frequently offset by a certain amount compared with the actual edge in the unfiltered image. According to the invention, the image matrix of the filtered image is therefore scanned in four different directions, namely column-by-column and line-by-line and in both diagonals, and local maxima and minima are determined in all these scanned pixel sequences. Between these maxima and minima, the actual grey scale variation in the unfiltered image is then approximated by a function having a turning point, for example a sine function, for example in accordance with the method of least error squares, and the turning point then specifies a point of an edge. Thus, interruptions in the edges and displacements compared with the actual image are avoided. <IMAGE>

Inventors

• 👤 Martin Walter Imme

Bibliographic Details