Application of gradient-based Hough transform to the detection of corrosion pits in optical images

In this paper, we introduce a circle detection technique named Hough transform to automatically recognize the corrosion pits in microscopic images. All the points in the input image are transformed into a parameter space, which is represented by a two-dimensional accumulative array with the same size of the original image. Local extreme values in the accumulative array, which represent the candidates of corrosion pits, are located using a maxima searching algorithm. The accuracy of detecting the number, radius and coordinate of pits from simulated images was examined. The results show that more than 95% of pits were successfully detected and the average errors of radius and coordinate are less than 10%, while these errors have negligible effect on the pit size distribution. The introduced method can also differentiate pits from scratches or inclusions, as indicated by the 100% accuracy of pit detection, from the simulated images presented in this study. Therefore, it is believed that the gradient-based Hough transform is a powerful method for the recognition of corrosion pits in microscopic images, making the statistical analysis of pit size and pit locations easier and more efficient.