Abstract
We present an efficient protocol for robust detection of 3D blobs from volumetric datasets. The approach has three steps. The first step of the process detects elliptic features by classifying the Hessian of the scale space representation of the volume data. These features are then grouped into 3D connected components, which are subsequently partitioned by computing a convex hull of each connected component. The proposed framework has been applied to a database of multicellular systems for detailed quantitative analysis.