ABSTRACT
Algorithms for subpixel edge localisation are becoming more popular as their goal is to provide super resolution. The proposed method for edge localisation does not use a numerical differentiation but enables minimisation of the digitising error and getting the subpixel coordinates of the edge point. The method is based on finding a shared point of intersection of two boundary curves, found as a result of low-frequency filtering. The low-frequency filtering is performed by means of a Gaussian filter with specific weighting coefficients. In a modern theory of digital image processing much attention is paid to the development of edge detection methods, because a detection of contrast differences is one of the main problems of image recognition in many technical vision systems. This is related to the intuitive understanding that the contour lines contain much information.
