This paper aims to develop an enhanced methodology for crowd counting through the integration of image super-resolution and density estimation techniques, addressing the critical challenge of accurate crowd analysis in low-quality images. Our proposed system combines Enhanced Super-Resolution Generative Adversarial Network (ESRGAN) and Congested Scene Recognition Network (CSRNet), creating a robust framework that can process images of varying quality and crowd densities. The model architecture allows for processing images of arbitrary size and quality, with ESRGAN’s upscaling mechanism particularly effective in enhancing detail preservation in crowded scenes. We conducted extensive experiments using the Shanghai Tech dataset, which includes diverse crowd scenarios and varying image qualities. Our comprehensive evaluation demonstrates the effectiveness of this hybrid approach in improving crowd counting accuracy, particularly in challenging low-resolution scenarios. The experimental results show that our integrated model achieves superior performance compared to traditional single-model approaches, advancing the field of crowd counting and contributing to public safety applications by providing more reliable crowd density estimates.