When processing photoelastic fringes, there exist several problems such as how to reduce unwanted noise generated by various causes, and how to extract accurately necessary fringes. Using modern techniques for image processing, experimental data are pre-processed in several steps such as enhancing, sharpening and filtering. Then binary and thinning processes are performed. Finally broken portions of fringes are connected manually to obtain smooth fringe patterns. However, determination of the threshold values for the binary process and the connection of broken fringe portions still pose various problems. It is also difficult to construct all of the fringes over the whole image area. In this paper, we would like to express the brightness distribution of fringe patterns of both isochromatics and isoclinics by the spline smoothing method on a microcomputer. Using this method, brightness distribution over the whole area can be successfully smoothed, and the location of extreme points that correspond to the fringes can be easily extracted. We would like to establish the “structure driven image processing method, ”as a valuable improvement for the automatic extraction of photoelastic fringes.