Hyperspectral images provide detail information because of their high spectral resolution. Having high spectral resolution makes them very useful for variety of applications. High spectral resolution requires very narrow wavelengths and neighbour bands which also brings some drawbacks like repetitive information and the burden of high dimensionality. Because of these drawbacks and the Hughes phenomenon the accuracy of classification decreases. In order to get rid of these disadvantages, dimensionality reduction of the data is advised. The purpose of this study is investigating the affect of dimensionality reduction methods on the classification accuracy of hyperspectral images together with the Gabor texture information in different scales and orientations. In this study first, the dimensionality of AVIRIS Indian Pine data is reduced to 20 with nine different methods. Second, Gabor features are generated using the first three PCAs of these images and then, these features are concatenated to these reduced images. Finally, both the reduced images and the Gabor features added images after reduction are classified with support vector machine. According to our results, Gabor features increase the classification accuracy of the data considerably. The data reduced with the Diffusion Maps Method and Gabor feature with 8 scales and 6 orientation achieves the maximum classification accuracy around 85%.