Contrary to the active detection methods, the passive sound source localization technologies only rely on the differences among the signals recorded by different receivers to localize the object generating the sound wave, without the knowledge of the sound source profile and excitation time. Time domain investigation methods have been widely applied in the analysis of sound source tracking, which greatly suffers from the high ambient noise problems. Instead, this paper applies an imaging method based on frequency domain cross-coherence (FDCC) operation to reconstruct the source to alleviate the effect of noise. The experimental results show the reliability of this method to recover the source in 3D. Moreover, a parabolic reflector is proposed to further increase the measurement amplitude and signal to noise ratio. The FDCC method successfully localizes the source even if the propagation path of the sound wave is disturbed by the reflector.