Improving the quality of extracted endmembers

Endmember extraction for spectral mixture analysis is a necessary step when endmember information is unknown. If endmembers are assumed to be pure pixels present in an image scene, endmember extraction is to search the most distinctive pixels. Popular algorithms using the criteria of simplex volume maximization (e.g., N-FINDR) and spectral signature similarity (e.g., Vertex Component Analysis) belong to this type. If pure pixel assumption is not imposed, endmember extraction usually is conducted by searching the signatures that can circumscribe the data cloud with the minimum volume. Both types of algorithms are affected by anomalous pixels since such outliers are very different from other pixels and act as interferers during simplex volume evaluation. In this paper, we propose a new approach that separates the endmember searching in normal and anomalous pixels. Real data experiments show that it can improve the quality of extracted endmembers.