Abundant information about the ground is contained in remote sensing images. However, due to bad weather conditions such as haze, they will have contrast degradation, color distortion, and detail loss, which will affect the implementation of subsequent computer vision tasks. Aiming at the problems that the existing dehazing methods for remote sensing images will cause unreal color restoration and blurred edges, this paper proposes a novel Multi-expert Collaborative Dehazing Network (MCD-Net). The network improves the dehazing performance by fusing the results of multiple experts. It uses the U-Net’s encoding architecture to extract image features, followed by multi-expert processing of edge and color information. Channel and pixel attention mechanisms further enhance channel and pixel features. Finally, it fuses multi-scale features to restore the image. Through quantitative and qualitative comparison with seven different dehazing methods on two public datasets, SateHaze1k and HRSD, MCD-Net achieved excellent restoration results, such as the highest PSNR (27.24dB) and SSIM (0.9017) on the HRSD-LHID, and has more natural color and higher definition. In addition, the rationality and importance of each network unit of MCD-Net are further verified by ablation experiments.