Hi,
When 2D-classifying small particles on crowded grids, the recentering mask often has two or more discontinuous regions, leading to suboptimal recentering behavior and loss of particles that might be retained if the convergence radius for recentering were larger.
It would be useful to have an option in Class2D to just keep the largest masked region (this would also help avoid including streaks or artefacts in the recentering mask).
