1990 Volume 30 Issue 7 Pages 482-489
A computer program for automated repair of defects in thinned grain boundary images has been developed. The basic data which is necessary to reproduce the geometrical shape of line images are the coordinates of characteristic points and the location of neighboring points connected by lines. Recognition of geometrical structures containing defects is done using the lists of characteristic points, from which a set of data of the distances and angles in the line image surrounding defects is prepared. Repair of grain boundary images consists of erasing undesirable lines and drawing lines between characteristic points. The repair operation involves rewriting of the contents of the data lists. The image reproduced from the repaired data lists is the network of triple points. The rules and method used in the repair operations are similar to those used during manual repair using visual observation. We found that line images containing more than 100 end points as defects could be repaired. With no abnormal structure in the input grain boundary image, the repaired image of the triple point network could be used for recognition of the shape of individual grains. The repair program could be used in a fully automated grain size analyzer.