<p>Restoring clear images from hazy images presents a significant challenge. Despite the significant advancements demonstrated by existing image dehazing algorithms, the effective extraction of features and the restoration of fine details from hazy images is still a challenging task. In this paper, we propose an end-to-end image dehazing algorithm based on deep learning and multiscale cross fusion. The haze-relevant features are extracted from different scales, and recursive residual network is employed to extract deep features while avoiding gradient diffusion. Information from various scales can be fused through cross-fusion process that incorporates an attention mechanism. The algorithm is independent from the atmospheric scattering model, and enables direct fusion of haze features across different scales, thereby mitigating the inaccuracies associated with traditional physical models and ultimately yielding a clear image. The experimental results on the RESIDE, I-HAZE, and O-HZAE datasets demonstrate that our algorithm achieves excellent performance in both synthetic and naturally hazy images, and our algorithm outperforms other comparison algorithms across several evaluation metrics.</p>