Balance problems such as inability to maintain stability and postural transition are common in patients with Parkinson’s disease (PD). In clinical, visual cues are used as rehabilitation program in improving postural stability. Ball catching movement is considered a kind of suitable training program for PD patients because of its characteristics of eyehand- foot coordination. Furthermore, studies have shown that optical flow could provide motion perceptions and improve movement performance of PD patients. The purpose of this study was to develop a 3D virtual reality (VR) training system for providing optical flow information during catching virtual balls under standing and one-step forward movements for subjects with PD. The arm-trunk movement and trunk movement were utilized as the assessment indices of balance and postural control by inertial sensors attached on waist and bilateral of wrists. The balance training system accompanied with evaluation system demonstrated that armtrunk control and balance on ball catching performance of PD patients could be improved with optical flow information.