This thesis presents a 3D stereoscopic interactive fish tank system that combines the 3D stereoscopy and ¡§controller-free¡¨ components. Based on the characteristics of human vision, when seeing the objects, the left eye image and right eye image will be slightly different, one can use the intensity information and the epipolar geometry to proceed matching, and then to generate the 3D depth information. This system allows a user to use gesture to do interaction. It estimates 3D objects depth information, and uses eyes distance, distance between the user and the sensor, disparity map to calculate the virtual objects¡¦ three-dimensional coordinate, and then transforms hand and virtual objects¡¦ coordinate into the same coordinate to allow accurate interaction. The system allows users to experience the innovative multi-media interactive entertainment.