Parkinson’s disease (PD) is one of the most common neurodegenerative disorders that causes postural instability and gait alterations, such as reduced walking speed, shorter step length, and gait asymmetry, exposing patients to a higher risk of falling. Recently, virtual reality (VR) was added to a treadmill, in order to promote motor functional recovery and neuroplastic processes. Twenty PD patients were enrolled and randomly assigned to two groups: the experimental group (EG) and the control group (CG). In particular, patients in the EG were trained with the C-Mill, an innovative type of treadmill, which is equipped with semi-immersive VR, whereas the CG performed conventional physiotherapy. Patients in both groups were evaluated through a specific motor assessment battery at baseline (T0) and after the training (T1). Comparing pre-(T0) and post-(T1) treatment scores, in the EG, we found statistical significances in the following outcome measures: 6 Minutes Walking Test (6MWT) (p < 0.0005), Timed up and go (TUG right) (p < 0.03), Berg Balance Scale (BBS) (p < 0.006), Tinetti Scale (TS) (p < 0.002), Falls Efficacy Scale- International (FES-I), (p < 0.03) Unified PD Rating Scale-III (UPDRS) (p < 0.002), and Functional Independence Measure (FIM) (p < 0.004). Also, the CG showed statistical significances after the training. Between-group (EG and CG) analysis showed significative statistical differences in 6MWT (p < 0.006), BBS (p < 0.006), TS (p < 0.008), FES-I (p < 0.01), and FIM (p < 0.009). From our results it emerges that both groups (EG and CG) achieved better outcome scores after the treatment, suggesting that both physiotherapy interventions were effective. However, the EG training using VR seemed to have induced more improvements, especially in gait and balance skills. Then, C-Mill could be a valid adjunctive treatment in the context of gait and balance disturbances, which are very common in the PD population.