In recent years, the field of storage and data processing has known a radical evolution, because of the large mass of data generated every minute. As a result, traditional tools and algorithms have become incapable of following this exponential evolution and yielding results within a reasonable time. Among the solutions that can be adopted to solve this problem, is the use of distributed data storage and parallel processing. In our work we used the distributed platform Spark, and a massive data set called hyperspectral image. Indeed, a hyperspectral image processing, such as visualization and feature extraction, has to deal with the large dimensionality of the image. Several dimensions reduction techniques exist in the literature. In this paper, we proposed a distributed and parallel version of Principal Component Analysis (PCA).