In order to solve the problem of obtaining clear images in foggy weather, this study conducted research on combining physical imaging models with Ostu threshold image segmentation algorithms, aiming to optimize computer defogging algorithms, improve the defogging processing performance and detail restoration ability of foggy images. The study analyzed the changes in transmittance of the research scene based on the atmospheric scattering model, and achieved image restoration of different fog concentration areas through image segmentation, thereby improving the defogging efficiency. The comparative test results show that the algorithm performs excellently in terms of convergence speed, fitness value, accuracy, and recall rate. In addition, the average errors of the model in the training and test sets were 1.35 and 0.98, respectively, which were smaller than those of other comparison algorithms, demonstrating superior analytical capabilities. The inference time of the proposed algorithm on the training and test sets was 0.97 s and 0.95 s, respectively, which was the fastest among all algorithms. At the same time, the algorithm processed 2.5 million samples on the training and test sets, using 124MB and 120MB of memory respectively, with floating-point operations of 8.4 GFLOP and 8.5 GFLOP, respectively. It demonstrated a good balance between efficiency and resource utilization. The experimental results of the defogging effect showed that the proposed algorithm outperformed other compared algorithms in terms of mean square error, peak signal-to-noise ratio, and structural similarity when processing foggy images in different scenarios, indicating that the algorithm could make the restored images closer to fog free images in different scenarios. Moreover, the image visibility and visual contrast of this algorithm were 87.83% and 86.52%, respectively, which were 37.48% and 46.19% higher than traditional algorithms. The experimental results have verified the effectiveness and feasibility of this method in practical applications, especially in autonomous driving, video surveillance, and remote sensing image processing, which have broad application potential and significant importance.