When analyzing stress distribution experimentally by the photoelastic method, fringe patterns may not be obtained over a whole area. It is necessary, for such cases, to reconstruct the fringe patterns by filling up the broken portions. Currently available techniques for this work are tedious and must be carried out by experts. This may be one of the reasons that the photoelastic method has not been accepted widely in industry. Various methods have been proposed to overcome this. Chief among them are; a hybrid analysis utilizing both numerical and experimental methods, and image processing of photoelastic fringe patterns. In the former, many problem still remain in comparing numerical results with experimental results. In the latter, not all of the procedures in such an analysis have been automated. This paper describes a newly developed algorithm for processing photoelastic fringe patterns on a microcomputer. Experimental data are represented by“smoothing spline functions. ”The locations of extreme points, fringes, and specimen contour are automatically extracted from these spline functions. In the smoothing process, breaks of fringe often occur over the region where the signal and noise ratio is small. Broken portions are reduced by superimposing several images on top of each other. The remaining broken portions, if exist, are automatically connected using locally applied planar curves.