are circled. This total LM set is far from being minimal. Actually, as shown in Fig. , only three of the disks are A two-dimensional binary object can be totally covered by a set of disks. From this follows that any object might be needed to cover the entire object.
represented by these disks rather than by the pixels. This paper
The problem of finding the minimal set of maximum deals with the problem of finding the most efficient version of disks (MSD) has been treated by several authors [3, 6]. this representation which is called the minimal set of maximum They have all used the approach indicated in this introducdisks (MSD). The proposed algorithm picks candidate disks tion, namely to find the candidate disks by detecting local from the local maxima in the Euclidean distance transform. maxima in a distance transform of the object. The crucial
Then a relation table for the pixel coverage of these disks is
and nontrivial problem is then to reduce this set of candiestabished, but only for the border pixels which makes the table dates to the minimal set in an efficient way.
size reasonable. Three basic table reduction steps are executed
It should be noted that the MSD could always be found to extract and include necessary disks to MSD while eliminating in an exhaustive search by testing all possible subsets of and excluding the unnecessary disks.