Contiguity-enhanced k-means clustering algorithm for unsupervised multispectral image segmentation

Publisher Website
Google Scholar
Abstract
The recent and continuing construction of multi- and hyper-spectral imagers will provide detailed data cubes with information in both the spatial and spectral domain. This data shows great promise for remote sensing applications ranging from environmental and agricultural to national security interest. The reduction of this voluminous data to useful intermediate forms is necessary both for downlinking all those bits and for interpreting them. Smart on-board hardware is required, as well as sophisticated earth-bound processing. A segmented image is one kind of intermediate form which provides some measure of data compression. Traditional image segmentation algorithms treat pixels independently and cluster the pixels according only to their spectral information. This neglects the implicit spatial information that is available in the image. We will suggest a simple approach - a variant of the standard k-means algorithm - which uses both spatial and spectral properties of the image. The segmented image has the property that pixels which are spatially continuous are more likely to be in the same class than are random pairs of pixels. This property naturally comes at some cost in terms o of the compactness of the clusters in the spectral domain,but we have found that the spatial contiguity and spectral compactness properties are nearly 'orthogonal', which means that we can make considerable improvements in the one with minimal loss in the other.

This publication has 0 references indexed in Scilit: