Haze degradation of near-earth remote sensing images due to atmospheric aerosol scattering is a significant challenge in remote sensing applications. It not only reduces image quality but also impairs the accuracy of downstream tasks such as target detection, environmental monitoring, and disaster response. The growing demand for high-quality remote sensing data has led to an increasing need for effective haze removal methods. Traditional approaches often struggle with complex atmospheric conditions and rely heavily on single visible-light images. With the advent of multispectral and polarization-based imaging technologies, new opportunities have emerged for more precise and effective haze removal by utilizing the distinct polarimetric characteristics of light. This enables better penetration of light through haze particles, improving the clarity and accuracy of images, especially in challenging conditions. The proposed dehazing method, which integrates visible and near-infrared (NIR) polarization spectral images, addresses these challenges by accurately modeling airlight radiance and polarization degree.The dehazing method proposed in this study, based on the fusion of visible and NIR polarization multispectral images, consists of two main steps. In the first step, the atmospheric light polarization state distribution model is utilized to estimate the atmospheric light intensity and polarization degree at an infinite distance. By using the relationship between the Stokes vector and the Mueller matrix, the polarization spectral images corresponding to the maximum and minimum light intensities are obtained. These images are then processed to generate the visible polarization spectral dehazed image. In the second step, the dehazed visible image is fused with the NIR image to further improve the image quality, resulting in a final dehazed image with rich details (Fig.1).In order to objectively evaluate the performance of the proposed image fusion dehazing method, several commonly used image quality assessment metrics were employed for quantitative analysis. Table 1 presents the comparative results of the dehazing methods, including SCHAUL, HE, FFA-Net, and FREDERIKE, based on different metrics. The results demonstrate that the proposed method excels in all evaluated metrics, particularly in Information Entropy (IE) and Contrast (IC), both showing significant improvements. Specifically, the information entropy improved by approximately 3.6% relative to the original image, while the contrast increased by about 3.26 times, indicating higher image information retention and better dehazing performance. Compared to other methods, the proposed method preserved more image details during the dehazing process, particularly in the handling of vegetation and distant scene areas. The method demonstrated superior dehazing effects, with a notable recovery of fine details in both the visible and near-infrared spectral bands. This reflects the method's ability to achieve photorealistic results that enhance visual clarity and retain structural details across different layers of the image, particularly in challenging conditions like haze. Furthermore, the adaptive fusion approach used in the proposed method played a key role in the effective combination of the visible and NIR polarization spectral images, leading to a significant improvement in both the detail recovery and overall image clarity. Despite the introduction of some granularity noise in certain high-frequency edge regions due to slight inconsistencies in pixel values during the fusion process, the method's ability to maintain the global image quality remains outstanding, with noise levels effectively reduced after post-processing. The results, therefore, not only validate the effectiveness of the method but also highlight its superiority over existing dehazing techniques.The proposed dehazing method, based on the fusion of visible and NIR polarization spectral images, demonstrates significant improvements in image quality. By leveraging both the visible and NIR spectral characteristics, the method effectively mitigates the haze and enhances the fine details of the images. The fusion of these two spectral sources allows for better retention of information and improved contrast. The results indicate that the method significantly outperforms conventional dehazing techniques in terms of both subjective image quality and objective evaluation metrics. The proposed approach not only restores image details but also preserves the essential information of the scene, making it a promising solution for remote sensing applications, particularly in hazy environments where traditional methods fail to produce satisfactory results. The effectiveness of this method is validated through both qualitative and quantitative assessments, confirming its potential for practical use in various imaging and remote sensing tasks.