ATLAS is a new spot detection method. The spots size is automatically selected and the detection threshold adapts to the local image dynamics.
ATLAS relies on the Laplacian of Gaussian (LoG) filter, which both reduces noise and enhances spots. A multiscale representation of the image is built to automatically select the optimal LoG variance. Then, local statistics of the LoG image are estimated in a Gaussian window, and the detection threshold is pointwise inferred from a probability of false alarm (PFA). Hence, the user only has to specify:
- the standart deviation of the Gaussian window,
- the PFA value.
The Gaussian window must be about the size of the background structures.
Increasing the PFA increases the number of detections.
You can try ATLAS on our mobyle web portal: Try ATLAS
An ImageJ plugin will be available soon.
A. Basset, J. Boulanger, P. Bouthemy, C. Kervrann, J. Salamero, SLT-LoG: A vesicle segmentation method with automatic scale selection and local thresholding applied to TIRF microscopy. Proc. IEEE Int. Symp. on Biomedical Imaging (ISBI’14), Beijing, April 2014.