Abstract: A method and device are provided for segmenting an image of pixels into a number of fields. First the method finds field separators using the background of the image, in particular white areas on a newspaper page. Based on the areas in the image, a graph is constructed that has edges corresponding to the white areas and vertices where vertical and horizontal white areas intersect. The segmenting starts with assigning weights to the edges, in particular a weight indicating the Euclidean distance between the vertices. Then a list of shortest cycles is constructed via the edges and vertices in the graph. The fields are defined by the vertices and edges of the shortest cycles of the list.

Abstract: A method and device are provided for segmenting an image of pixels into a number of fields. First the method finds field separators using the background of the image, in particular white areas on a newspaper page. Based on the areas in the image, a graph is constructed that has edges corresponding to the white areas and vertices where vertical and horizontal white areas intersect. The segmenting starts with assigning weights to the edges, in particular a weight indicating the Euclidean distance between the vertices. Then a list of shortest cycles is constructed via the edges and vertices in the graph. The fields are defined by the vertices and edges of the shortest cycles of the list.