Face recognition technology has greatly improved, but identifying faces with occlusions is still difficult due to partial obstructions that change key facial features. This study aims to enhance occluded face recognition by integrating different techniques. We propose two approaches: the first integrates texture analysis methods such as Local Binary Patterns (LBPs), Local Ternary Patterns (LTPs), and Completed Local Binary Patterns (CLBPs) with Convolutional Neural Networks (CNNs) and employs the softmax function for more accurate classification. In the second approach, we combine these texture analysis methods with dimensionality reduction methods like PCA and LDA, and classification algorithms like SVM. Additionally, the Histogram of Oriented Gradients (HOG) is used as a feature extraction step in both approaches to improve resilience to variations in occlusions. We evaluate entropy for each block size to identify the most informative features that ensure the blocks with the highest entropy contain the most variation and useful information. By applying all techniques to each non-overlapping block, we optimize feature extraction and handle various occlusion patterns. This multi-faceted approach allows for advanced feature extraction and greater compatibility with a wide range of occlusion patterns. Our evaluation using the AR dataset, known for its realistic occlusion conditions, demonstrates vital improvements in recognition performance. These results highlight the robustness and effectiveness of our integrated approach to advancing occluded face recognition technology.