EDGE EXTRACTING PROCESSING METHOD

摘要

PURPOSE:To reduce the position error caused by a filter of a large scale and to effectively extract an edge by integrating the information of different scales. CONSTITUTION:An input gradation image 1 obtained from a digital gradation image is supplied to an edge extracting part 2. A large area information extracting pert 4 of the part 2 carries out the zero crossing and the convolution operation with a Laplacean Gaussian filter of a large scale against the entire input gradation image. Thus a rough shape of an edge is obtained. Then a detailed information extracting part 5 carries out a convolution operation with a Laplacean Gaussian filter of a small scale and the zero crossing against the vicinity of a noted point on the obtained rough shape of the edge. Thus the details of the edge are extracted and an output edge image 3 is obtained. In such a way, the rough shape of an edge is obtained with use of a filter of the larger one of two scales sigma and a filter of a small scale is applied to a nearby point in a tracking mode of the rough shape of the edge.