Landscape Image Defogging System Based on DCPA Algorithm Optimization

Abstract

As the global climate environment deteriorates gradually, the collected images are covered by fog, which reduces the clarity of the images. Therefore, the processing of fog images is very important. In landscape image defogging research, the defogging process may be affected by factors such as data quality, noise interference, and computational efficiency. To improve the defogging effect of landscape fog images, a landscape image defogging system was put forward with the optimization of dark channel prior algorithm. The image defogging algorithm was combined with the improved atmospheric scattering model estimation algorithm and the dark channel convolutional network image defogging algorithm to achieve image defogging. The atmospheric light estimation method based on atmospheric scattering model combined transmittance map and grayscale map information to achieve optimization of defogging effect. In the improvement of the dark channel prior algorithm, a convolutional network was introduced for feature extraction to enhance the smoothing of image brightness changes and transmittance estimation. The research findings demonstrated that the signal-to-noise ratio of the image defogging algorithm estimated by the atmospheric scattering model could reach up to 19dB, which was about 15.4% higher than that of existing 5 image defogging algorithms on average, indicating that the image resolution of the algorithm was higher after defogging. In the Reside dataset, the image defogging algorithm based on dark channel prior increased the signal-to-noise ratio by about 9.5%, the average gradient by about 10.4%, the structural similarity by about 12%, and the information entropy by about 5.8%, indicating that the effect of the algorithm was stable and the image defogging effect was good. The dark channel convolutional network image defogging algorithm had less running time and reduced the complexity of the defogging structure, by contrast, it reduced the running time by about 67%. The average scores for the operability, stability, and defogging effect of the system were 9.87 points, 9.85 points, and 9.54 points, respectively, indicating good performance of the system. The user feedback on natural landscape fog maps, architectural landscape fog maps, and historical landscape fog maps is good, and the user experience is high.