| Abstract: | In this research, a measure of robustness for edge detectors is developed. This measure evaluates quantitatively the robustness of any chosen edge detection algorithms. In this research, edge detectors: Argyle, Macleod, LoG[the Laplacian of Gaussian), and DoG(the first derivative of Gaussian) for digital image processing are evaluated in terms of their respective performance. The measure consists of two measures: one is for the performance and the other is for the sensitivity of the performance. The measure is applied by comparing the chosen edge detectors with a chosen test image: chart. Several different levels of the Gaussian random noise are added to the test image in order to compare the performance under noise. The size of the edge operators is changed to study the tradeoff relation between the peformance and the sensitivity. Based on these two variations: the noise level and the operator size, the total performance and the total sensitivity are measured, which reflects the total robustness in the performance. |
|---|