We report a discovery of a companion candidate around one of {it Kepler} Objects of Interest (KOIs), KOI-94, and results of our quantitative investigation of the possibility that planetary candidates around KOI-94 are false positives. KOI-94 has a planetary system in which four planetary detections have been reported by {it Kepler}, suggesting that this system is intriguing to study the dynamical evolutions of planets. However, while two of those detections (KOI-94.01 and 03) have been made robust by previous observations, the others (KOI-94.02 and 04) are marginal detections, for which future confirmations with various techniques are required. We have conducted high-contrast direct imaging observations with Subaru/HiCIAO in $H$ band and detected a faint object located at a separation of $sim0.6$ from KOI-94. The object has a contrast of $sim 1times 10^{-3}$ in $H$ band, and corresponds to an M type star on the assumption that the object is at the same distance of KOI-94. Based on our analysis, KOI-94.02 is likely to be a real planet because of its transit depth, while KOI-94.04 can be a false positive due to the companion candidate. The success in detecting the companion candidate suggests that high-contrast direct imaging observations are important keys to examine false positives of KOIs. On the other hand, our transit light curve reanalyses lead to a better period estimate of KOI-94.04 than that on the KOI catalogue and show that the planetary candidate has the same limb darkening parameter value as the other planetary candidates in the KOI-94 system, suggesting that KOI-94.04 is also a real planet in the system.