In this paper, we introduce a new clustering algorithm called Improved Kernel Possibilistic
Fuzzy C-Means algorithm (ImKPFCM), based on the kernel method and possibilistic approach.
The proposed ImKPFCM algorithm corrects several FCM, PFCM and GPFCM algorithms shortcomings,
reliably detects clustering centers and allows in addition to use Euclidean distance,
the employment of other more powerful additional norms able to handle various complex
situations. In this study, we applied ImKPFCM algorithm as a new image clustering
method on the basis of Tchebychev orthogonal moments to extract feature vectors and
then compared it with FCM, PFCM and GPFCM algorithms to evaluate its performance.
The comparative study results applied to several image dataset, revealed that the
ImKPFCM clustering algorithm improves the clustering accuracy over the FCM, PFCM and
GPFCM methods. Therefore, we conclude that the ImKPFCM algorithm is more efficient
and produces satisfactory image clustering results.
