Abstract: Morphological operators used in the automated morphological endmember extraction (AMEE) algorithm didn't acquire correct result in the area of pure pixel concentration distribution. The dilation operation only chose one candidate endmember from each structure element and would lose some important pixels. In order to solve the problem, the AMEE algorithm was modified by an improved morphological operator and new structural element. The improved morphological operator was proposed after introducing the concept of reference spectral vector, and a new calculation method of morphological eccentricity index was given. To avoid information loss, four candidate endmembers were chosen from each improved even-numbered structure element. The modified automated morphological endmember extraction algorithm was tested based on a hyperspectral data set. The experimental results show that the improved method can obtain correct candidate endmembers from the area of pure pixel concentration distribution, and information loss in the procedure of dilation is also avoided. The proposed method produces more accurate results of endmember extraction and the spectral unmixing.