A Kind of Sea-fog Removal Algorithm Fusing Image Enhancement and End-to-end Network

Generally, for images with sea fog, the distribution of fog density is uneven, and the whole scenes tend to the darkness, if only the image enhancement algorithm is used to defog, it could cause color distortion of the overall image, and the details blurring of the target image; the end-to-end defogging algorithm has obvious defogging effects for the uniform mist image on the land, but it is not good for sea fog images with dense fog and dark scenes. In this paper, a kind of single sea fog image defogging method is proposed, which fuses the image enhancement with end-to-end network. Firstly, the image enhancement based on Multi Scale Retinex was adopted, to improve the contrast and brightness of the uneven fog image; Then, the end-to-end network was used to extract the transmittance, and then the transmittance was optimized by the guided filter; Finally, through removing fog based on the atmospheric scattering model, the edge blur and color distortion of the object are improved. We had experimental analysis compared our proposed method with several mainstream defogging algorithms, to evaluate the defogging quality from both subjective and objective aspects separately, the experimental results show that the proposed single-image sea fog removal algorithm has obvious defogging effects and good image restoration ability.