Nevertheless recent development in hardware devices has been remarkable, it is still difficult to get noise-free data of photoelastic fringes with a wide use polariscope. It must be reasonable, moreover, to consider that photoelastic data include both dense and sparse portions of fringe. Thus, unavoidable noises affect the accuracy of analysis for the location of fringes, particularly in highly dense portions. In order to improve the accuracy of analysis, an atempt of linearly proportional enlargement of the dense area seems to be available. This method, however, could not be useful, because of several kinds of aberration of lens and discrepancy in a scale-down of image into the original size. This paper describes the method of conversion of a fringe pattern obtained in a rectangular array (Cartesian coordinate) into a polar coordinate. After analyzing locally enlarged portions, fringes are transformed inversely back into the original coordinate system. Several examples of application of this method show its availability in extraction of photoelastic isochromatic fringes.