In the field of face recognition, particularly within surveillance systems, accurately identifying faces in low-resolution images poses a significant challenge. This study introduces an advanced algorithm designed to enhance face recognition accuracy by mitigating the effects of low image resolution. The proposed method encompasses a multi-step process: initially, high-resolution facial images are down-sampled to emulate low-resolution conditions. Following this, three interpolation techniques—nearest neighbor, bilinear, and bicubic are utilized to upscale these low-resolution images, aiming to recover and enhance lost details. For feature extraction, the study employs a combination of Scale-Invariant Feature Transform, Speeded-Up Robust Features, Local Binary Patterns (LBP), and Block-Based Discrete Cosine Transform. These techniques are assessed for their efficacy in representing facial features under various conditions. The extracted features are then classified using a range of methods, including Support Vector Machine, MobileNetV1, MobileNetV2, Convolutional Neural Networks (CNN), Artificial Neural Networks (ANN), and hybrid deep learning models. The performance of the recognition system is evaluated based on identification rates across different resolutions (e.g., 112 × 92, 56 × 46, 92 × 112), interpolation methods, feature extraction techniques, and classification approaches. The results indicate that interpolating low-resolution images substantially improves recognition performance, with optimal feature extraction and classification combinations further enhancing accuracy. This comprehensive approach provides a robust framework for improving face recognition systems in low-resolution scenarios, offering significant advancements for surveillance and security applications.