Pyramid-based image segmentation using multisensory data

摘要

A new approach is developed for segmenting scenes using multisensory data. The approach is based on the pyramid data structure. Image pyramids are built for each sensing modality. Information fusion between these pyramids is used to establish the scene segmentation. The application of this approach to thermal and the visual images of scenes is discussed. The advantages of this approach over previous techniques which use a single imaging modality are discussed. Experimental results using real multisensory data are presented to illustrate the performance of the new technique. The segmentation provided by this approach is shown to be suited for use by vision systems which classify objects (image regions) using multisensory data.