In contrast to techniques currently used such as the fundus camera, the strain to which the patient is exposed during the generation of an image of the ocular fundus is reduced by moving a laser beam across the retina in the form of a scanning raster and by directing the light reflected from the retina through a photoelectric receiver to generate a television image. To achieve a very high resolution, the scanning operation provides at least in the interesting image parts signals which are representative of a greater number of image points than corresponds to the television standard. These image parts are selected sufficiently large to have each point of the raster assigned a separate image signal value when displayed on the television screen. During the time the scanning beam moves within the frame of a selected image part, its intensity is increased for improved image contrast. The apparatus for implementing this image-forming method utilizes at least one polygonal scanner and includes an active image element that cooperates with a sensor to form a closed loop circuit for optical focusing of the image under adaptive control. An electronic shutter is provided to control the intensity of the scanning beam. In addition to the formation of images, the apparatus also permits a measurement to be made of the spatial blood flow distribution in the fundus and of the degree of oxygen saturation of the blood in the retina of the eye under observation.