A tensor approach for local structure analysis in multi-dimensional images

Using the n-dimensional structure tensor, linear symmetries within local neighborhoods of multi-dimensional image structures can be detected. Depending on the image content this information yields a a variety of features, such as velocity, local orientation, texture and structure information that can be used for visualization. The proposed technique is well suited to be applied within a
multi-resolution framework as well as for multi-component images and is not restricted in the dimensionality of the image content. However, the structure tensor requires a careful interpretation with increasing dimensionality. In this paper the implementation of the tensor approach will be introduced and the application to 3D image analysis and motion estimation in in spatio-temporal images will be discussed in detail.