In the literature, some labeling algorithms of nxn binary image, which are generally
claimed as optimal, are only optimal with regards to the algorithmic aspect but not to the
architectural aspect. Aside from this, the complexities obtained are not pure because the constant
still depends on the value of n. This paper presents a labeling algorithm, which reaches purely
Processor-Time optimal performance. This means that the optimal performance is not only
reached considering its algorithm but also its architecture and “pure” means that the complexity
obtained does not depend on the value of n. The algorithm complexity obtained is O(cn) using
O(n) number of processors. In this paper the justification of complexity obtained is given and the
performance comparison with other existing algorithms is discussed.