Synthesising objects and scenes using the reverse distance transformation in 2D and 3D

Abstract

The reverse distance transformation has proved useful in image synthesis. This paper describes how digital objects are created from a number of seed labels in an image. The shape of the obtained objects depends on the metric used. In 2D the Euclidean and the 3-4 metrics are mentioned, and in 3D the D6, the D26, and the 3-4-5 metrics are discussed. The proposed method has no need of expensive CAD systems. It is an excellent image synthesising tool when developing image processing algorithms, i. e. shape quantification, visualisation, scene analysis and range imaging, as the obtained objects are well-defined in the image. The method is most advantageous in 3D, as there is an increasing need for volume images, but synthesising objects in 2D can also be useful.