The invention relates to a method and a device for improving the depth impression of stereoscopic images and image sequences. In autostereoscopic multi-viewer display devices, generally a plurality of intermediate perspectives are generated, which lead to a reduced stereo base upon perception by the viewers. The stereo base widening presented in this application leads to a significant improvement and thus to a more realistic depth impression. It can either be effected during recording in the camera or be integrated into a display device. The improvement in the depth impression is achieved by the generation of synthetic perspectives situated, in the viewing direction of the camera lenses, on the left and right of the extreme left and extreme right recorded camera perspective on the right and left lengthening of the connection line formed by the extreme left and extreme right camera perspectives. These synthetic perspectives are calculated only on the basis of a disparity map, which is supplied or which is calculated in a preprocessing step. In this case, the method presented solves the following problems: 1. Calculation of the new extension perspectives, 2. Correct repositioning of the camera perspectives supplied within the visual zones, 3. Definition of which disparities are intended to be continued in the case of collision in the extension and 4. Interpolation of the image regions which become visible as a result of the extension of the stereo base. In this case, instances of left and right masking are also identified and optically correctly maintained and supplemented in the extension perspectives.