Effective Hyper-Spectral Image Segmentation Using Multi-scale Geometric Analysis

The wide availability of multispectral images has
fostered the development of new algorithms for remote sensing
applications. These applications range from agricultural and
environmental to military use. Nevertheless, the analysis of
such voluminous data requires advanced analysis and
computational methodologies as well as advanced hardware
and computational methods. In this paper we introduce a new
state-of-the-art method for segmentation of Hyperspectral
images which uses both spectral and spatial information
simultaneously. The proposed methodology is based on a multiscale geometric transformation, called the Beamlet Transform,
and the Beamlet Decorated Recursive Dyadic Partitioning (BDRDP). The method is applicable for both mono-spectral and
multispectral images where each pixel has its corresponding
spectral profile vector.
The proposed segmentation method is especially effective
when the underlying image consists of relatively large segments
with smooth boundaries. In this case, it performs exceptionally
well even when the Signal to Noise Ratio (SNR) is extremely
low. The method is unsupervised and assumes no prior
knowledge of the image characteristics or features.
Furthermore, it involves a free sensitivity parameter which
allows fine tuning for a specific application, and thus improving
segmentation results. Despite of being relatively complex and
sophisticated, the proposed segmentation algorithm has a low
computational complexity of . This is achieved by
implicit computations through the Pseudo-Polar Fast Fourier
transform (PPFFT). In order to validate the efficiency of our
method, we have used the Lark algorithm which also combines
spectral and spatial analysis but lacks the multi-scale property,
for segmentation of multi-spectral images and compared its
performance to the method proposed in this paper. These
comparisons showed that our new proposed method outperforms the Lark algorithm and emphasized the effectiveness
of multi-scale analysis. The proposed method was successfully
applied to real aerial multi-spectral imagery for the application
of estimating nitrogen levels in agricultural areas.