Stereo disparity is a major factor affecting visual comfort in stereoscopic video, and human eyes will feel uncomfortable when viewing stereoscopic scenes with large disparities. Therefore, in order to improve stereoscopic visual comfort, disparities was adjusted by a linear and a nonlinear methods, and the excessive disparities were reduced. Based on the stereo display parameters and human physiological parameters, the disparity range of the Panum's fusional area for human eyes was calculated, and the overall range of stereo video disparities were linearly adjusted to this range. Furthermore, based on the visual attention mechanism and motion information in the stereoscopic video, nonlinear disparity remapping was performed on the significant area of the video, and new stereo frames were synthesized using virtual viewpoint synthesis techniques based on the adjusted disparity maps. Finally, subjective experiments were conducted to evaluate the viewing comfort of original and the new stereoscopic videos. Experimental results has shown that the comfort values of new videos have been improves. It indicates that the proposed algorithm can adjust disparities effectively and improve the stereoscopic visual comfort by generate new stereoscopic videos.