Flexible rhino endoscopes have been extensively used in many surgical and diagnostic procedures, such as in a rhinoscopy to examine the tumor or polyp in nasal cavity. However, the orientation of flexible endoscopes is only based on the endoscopic images, making its manipulation complex and requiring clinical experience. Therefore, we proposed a hand-held manipulator to motorize a standard flexible rhino endoscope. In this paper, we present a method to steer the motorized flexible rhino endoscope by using endoscopic images. A vision technique is used to extract the dark region from the image, which represents the distant inner space of nasal cavity and indicates the heading direction of the endoscope. Using this computer recognized target orientation, automatic endoscope steering becomes possible. Experiments were conducted in an anatomical model to simulate rhinoscopy. The results show that the image-guided steering is significantly faster than traditional manual steering and that the measured accuracy of the system fulfills current clinical requirements.