Non-negative Matrix Factorization with Sparseness Constraints
The Journal of Machine Learning Research
| Hi-index | 0.00 |
In the hyperspectral analysis, the spatial correlation information is potentially valuable for hyperspectral unmixing. In this paper, we propose a new model, denoted "kernel spatial complexity-based nonnegative matrix factorization" (KSCNMF), to unmix the nonlinear mixed data. The method is derived in the feature space, which is kernelized in terms of the kernel functions in order to avoid explicit computation in the high-dimension feature space. In the algorithm, input data are implicitly mapped into a high-dimensional feature space by a nonlinear mapping, which is associated with a kernel function. As a result the high order relationships and more useful features between the spectral data can be exploited. Experimental results based on a set of simulated data and a real hyperspectral image demonstrate that the proposed method for decomposition of nonlinear mixed pixels has excellent performance.