Spatially-Weighted Sparse Coding

Spatially-Weighted Sparse Coding for Hyperspectral Image Classification

A hyperspectral image is a collection of pixels that represent a given scene or object, where pixels represent the reflected solar radiation from the Earth’s surface in many narrow spectral bands. At each pixel, the spectral features form a vector whose elements correspond to the narrow bands covering visible to infrared regions of the spectrum.

The high spatial and spectral resolution of a hyperspectral image provides the potential for each pixel to be accurately and robustly labeled as one of a known set of classes. Hyperspectral image classification has been applied to both urban and  agricultural scenery. Various methods have been developed for this application.

In this work, a spatially-weighted sparse unmixing approach is proposed as a front-end for hyperspectral image classification using a linear SVM. The idea is to partition the pixels of a hyperspectral image into a number of disjoint spatial neighborhoods. Since neighboring pixels are often composed of similar materials, their sparse codes are encouraged to have similar sparsity patterns. This is accomplished by means of a reweighted l1 framework where it is assumed that fractional abundances of neighboring pixels are distributed according to a common Laplacian Scale Mixture (LSM) prior with a shared scale parameter. This shared parameter determines which endmembers contribute to the group of pixels. Experiments on the AVIRIS Indian Pines show that the model is very effective in finding discriminative representations for HSI pixels, especially when the training data is limited.

Related Publication:

  • Soltani-Farani A., and Rabiee H. R., “Spatially-Weighted Sparse Coding for Hyperspectral Image Classification”, IEEE Geoscience and Remote Sensing Letters, vol. 12, no. 1, pp. 107-111, Jan. 2015

Project code:

  • This code is written for MATLAB and contains routines for several hyperspectral image classification methods used in the above publication. Please cite the above work if you use this software and contact first author in case of any problems.

People involved:

  • Ali Soltani-Farani, Hamid R. Rabiee

 

Sparse Signal Processing Group