The result of feature extraction on multi-spectral image is directly related to the complex degree of target recognition algorithm. And the performance of final target recognition is also influenced. The traditional local binary pattern (LBP) feature extraction method is researched. And several new LBP feature such as directional LBP feature describer, adaptive LBP feature describer and directional adaptive feature describer are introduced. And they are applied to texture features extraction of multi-spectral images. Experimental results show that LBP feature describer and its improvement methods are very suitable for extracting the texture features of multi-spectral images.