Edge detection by regularized cubic B-spline fitting

In this paper, regularized cubic B-spline fitting (RCBS), a technique which combines conventional cubic B-splines with regularization techniques, is proposed for the detection of edges. In addition, a quantitative performance evaluation model, called the false-detected to correct-detected ratio (FCR) model, is proposed for the evaluation of edge detectors. The FCR model attempts to provide a common ground for the quantitative performance evaluation and comparison between different edge detection schemes. The experimental results using the FCR model show that the RCBS operator is superior with respect to noise immunity and localization in comparison with two commonly used operators: the Sobel operator and Haralick's bivariate cubic polynomial operator.<
>