Teleoperation of humanoid robots for disaster response tasks, controlled by a master-slave system using a head-mounted display, requires a detailed frame of view for precise operation. We previously investigated an automatic zoom method, and images zoomed when the robot arm entered the work area could improve teleoperation performance. However, automatic zooming failed to consider the operator's intentions, including zoom levels and timing, causing high mental workload and motion sickness for operators. Therefore, this study proposes a zoom method with a low mental workload and reduced level of motion sickness. Motion sickness is less likely to be induced under active conditions, wherein the operator can manually control the image. Moreover, upper limbs should not be used to operate the zoom feature because such use can disturb the master-slave operation. Furthermore, a zoom method should reflect the operator's intentions, including zoom levels and timing. Thus, we develop a zoom method based on head movement, which can be regarded as typical gaze movement. The results of experiments using a real robot system indicated that the proposed method could reduce the mental workload and work time without inducing motion sickness, compared with zoom operation using a foot pedal interface.