In this proposed method, we used a new neighborhood shape for obtaining LBP as its representation of texture is superior to traditional LBP. We compute the proposed LBP on each non-overlapping blocks of a face image and a histogram of these LBPs is computed. We used these histograms as facial feature vectors for gender classification as these histograms shown their robustness to compression and uniform intensity variations. The classification task has been achieved by using Support Vector Machine (SVM). We compared our method with existing gender classification methods based on LBP with classifier being the same as SVM. The proposed LBP based descriptor outperforms the traditional LBP based methods. The system proposes a new approach to dense feature extraction for face recognition. This approach is Learning Compact Feature Descriptor and Adaptive Matching Framework. This approach consists of two steps. First, an encoding scheme is devised that compresses high-dimensional dense features into a compact representation by maximizing the intrauser correlation. Second, we develop an adaptive feature matching algorithm for effective classification. This scheme effectively converts high-dimensional dense features into a much more compact representation. Furthermore, we propose a new face matching method, called the ‘Adaptive Matching Framework’, and conduct experiments in four different face recognition scenarios: face recognition in the wild, aging face recognition, and matching near-infrared face images and optical face images, and the FERET test.