Automatic Detection of Solar Filaments Versus Manual Digitization

The technique which is presented here is based on careful attention to image cleaning in order to achieve robust automatic filament detection and avoid confusion between filaments and image defects, due to particular shapes of filaments. The main part of the detection process is based on seed selection and region growing. The procedures developed have been tested on four months of full-disk images from the Meudon Observatory. The results are compared with those manually generated in Meudon, for several hundred filaments. A very good correspondence is found, showing the robustness of the method described.