The most important ways to extract information from remotely sensed data is the classification methods, which can be divided into two groups of supervised and unsupervised techniques. Supervised methods need training data, which are costly and time-consuming; in contrast, unsupervised methods are automatic and rely solely on image data. Unsupervised classification methods are usually less accurate than supervised ones, but the fact that they require much less time and resources has put them at the center of attention. Clustering methods are among the most importance subcategories of unsupervised classification methods, and one of the most widely used techniques in this subcategory is the c-means method. The fuzzy version of c-means is the Fuzzy c-means (FCM) model, which is among the best-known clustering methods. But FCM is very sensitive to presence of outliers in data, and various revisions including Possibilistic c-means (PCM), Fuzzy Possibilistic c-means (FPCM) and Possibilistic Fuzzy c-means (PFCM) have been presented to solve this problem. PCM was the first method proposed for this purpose, but it led to problem of overlapping clusters; FPCM was then developed to overcome this issue, but it proved to be ineffectual for large datasets. Ultimately, PFCM model providing more flexibility than its predecessors managed to alleviate these issues. PFCM clustering algorithm is a combination of FCM and PCM that lacks the limitations of these methods and the FPCM algorithm. In this paper, one of the most successful versions of c-means, namely PFCM clustering algorithm, is used for classification of Hyperion hyperspectral image data, and the results are compared with the results of FCM algorithm. Experiments carried out in this study show that FPCM algorithm increases the overall accuracy by about 3%.