| Abstract [eng] |
One hundred various pictures with different size and number of elements were chosen for the method research of image outline evaluation. All these pictures were converted into grayscale pictures. Most of edge detection methods (filters) required to be blurred to reduce noise. Eight the most popular methods were chosen to evaluate the image outline: Canny, Sobel, Prewitt, Roberts, Zerocross, Laplacian, LoG, Marr-Hildreth. A Root Mean Square Error (RMSE) was computed for each edge picture with the best-chosen outline. |
