Image defogging algorithm based on multi-scale residuals and coordinate space attention

The rapid development of intelligent algorithms has provided strong support for computer vision tasks. Image dehazing, as a core part of the preprocessing stage of the visual system, is of vital importance for the reliability of downstream tasks. The existing methods for removing fog all have problems such as insufficient capture of fog characteristics and easy loss of details in key areas. Therefore, the study proposes an image defogging algorithm based on multi-scale residual network and coordinate space attention. Experiments show that the feature extraction module of this algorithm has a root mean square of 0.093 in the high concentration fog region, which can effectively strengthen key features and suppress redundant information. In addition, the structural similarity index of the algorithm reaches 0.9738, which is improved by 0.0471 compared with the comparison algorithm, and the peak signal-to-noise ratio reaches 29.37dB, which is an obvious advantage compared with the comparison algorithm. The above results show that the algorithm proposed in the study is able to achieve efficient defogging processing of foggy images, and the generated fog-free images have natural color and clear texture, which provides a new path for the defogging method of images.