Jump label

Service navigation

Main navigation

You are here:

Main content

Generation of limited depth-of-field using digital signal processing

The advantages of digital picture recording are obvious and convince home users as well as professional cinematographers. Professional film production has high requirements concerning the possibilities of art work. A very important part is the use of a limited depth-of-field to focus on important picture elements - a well known and very often used effect in movies. Using this effect enables the cinematographer to give a depth information to the viewer.

ja_antike
Image with large depth-of-field

This limited depth-of-field is a property of "analogue" 35mm film cameras due to its long focal length. It is one of the most important elements of the so called "filmlook". Modern HD-cameras have - in comparison to the 35mm film cameras - only a small image sensor. Therefore a smaller focal length is needed to get the same field of view. Because of this short focal length digital cameras have a wide depth-of-field. It is not always possible to compensate this by selecting a smaller f-stop and so the effect of limited depth-of-field is restricted.

The purpose of the research is to find an other way to accomplish the effect of limited depth-of-field. Therefore the scene is recorded with a wide depth-of-field and the desired effect is added during the post processing phase of the production. Thus the cinematographer has more liberties in generating the depth-of-field effect than by using a 35mm film camera.

ja_antikedepth
Depth map with the selected depth-of-field

 There are some algorithms for generating the desired depth-of-field effect in synthetic images. They are often used in computer games on modern graphics cards. In behalf of a fast computation some picture errors are accepted. Other algorithms, developed for render software, can not be used here because the renderer has image and scene information which are not available on recorded pictures. The renderer has information about hidden image parts, which contribute to the later filtering for the generation of the limited depth-of-field. This hidden image information has to be reconstructed from the surrounding image information.

ja_antikefiltered
Generated limited depth-of-field

In addition to this a depth map from the image is needed. When using synthetic images this information can be generated by the renderer. For real images this information has to be obtained during picture recording.



Sub content

Contact

Prof. Dr.-Ing. R├╝diger Kays
Tel.: 0231 755-2100