Face Recognition has become the challenging and interesting research topic in the last few years. The aim is to design a robust Face Recognition System under different environmental conditions like illumination, pose and occlusion. These are the three major challenges in Face Recognition which may hinder the Face Recognition system. By combining the three successful representations such as Gabor filters, CS-LBP and TLPP better performance can be achieved as compared to just considering them individually. CS-LBP is used for describing interest regions which have good tolerance to illumination and computational efficiency and TLPP is used to take the data directly in the form of tensors as input. Since the number of the combined feature sets are more only a few feature sets is selected to be trained by the Support Vector Machine classifier. A number of experiments are conducted using YouTube celebrity, McGill Face dataset and as well as the own collected sequence under different conditions such as illumination variations, different poses, occlusion including indoor and outdoor scenes. This approach provides better results compared to traditional approaches.