Exoplanets detection by direct imaging remains one of the most challenging field of modern astronomy. The signal of the star can prevent the detection of orbiting companions in single datasets, but combining information from several observations helps boost the detection limits. We propose a new algorithm named PACOME, based on PACO’s approach, which optimally combines, in a maximum likelihood sense, multi-epoch datasets and improves the detection sensitivity of potential exoplanets by taking into account their orbital motions. The efficiency of the algorithm is tested on the well-known exoplanetary system 51 Eridani.