Currently we implemented DBSCAN for clustering the masked data in RGDR. However, DBSCAN can be very memory and computationally intensive. This is also because DBSCAN is originally intended to cluster point cloud data, instead of a fixed grid.
To this end I was thinking about alternative clustering methods we could implement. One of these could be a "flood fill" algorithm, like the paint bucket in MS Paint. One example of this is in scikit-image.
Currently we implemented DBSCAN for clustering the masked data in RGDR. However, DBSCAN can be very memory and computationally intensive. This is also because DBSCAN is originally intended to cluster point cloud data, instead of a fixed grid.
To this end I was thinking about alternative clustering methods we could implement. One of these could be a "flood fill" algorithm, like the paint bucket in MS Paint. One example of this is in scikit-image.