It would be more efficient to do something computational based on the pixels.
If you are willing to indulge in some custom OpenGL or DX coding you could draw all of the larger circle parts as filled circles in an OR logic operation, then use the resulting image to create some kind of outline of those shapes, perhaps by jittering or something, and then finally draw the centers of the objects.
If you are willing to indulge in some custom OpenGL or DX coding you could draw all of the larger circle parts as filled circles in an OR logic operation, then use the resulting image to create some kind of outline of those shapes, perhaps by jittering or something, and then finally draw the centers of the objects.