Abstract: The appearance of foggy weather makes the collected images have low contrast and weak identify ability. In order to improve the functional stability of the vision system, an image defogging algorithm was proposed in this paper. To solve the halo problem existed in the dehazing image with the current popular dehazing algorithm, a morphological cascade algorithm was proposed to estimate the scene transmittance. On this basis, in order to make the scene transmittance keeping clear edge details and maintaining image smoothness, a weighted least square method was introduced to refine the transmittance. And then, considering image facing highlights and large areas of bright white, an improved hierarchical search algorithm was used based on the quad-tree method to accurately estimate the atmospheric light value in the sky area. Finally, the atmospheric scattering model was introduced, and the calculated transmittance and atmospheric light value were input the model to obtain dehazing image. Comparing the results with that of the current popular defogging algorithms qualitatively and quantitatively, the proposed method show that it can provide a better recovery effect in preserving depth edges, color quality and details, getting more clearly in-depth details, color quality and more natural. The proposed algorithm was tested on commonly used dehaze images and O-haze datasets. The test results show that the mean square error reduction can reach up to 7.0%, the signal-to-noise ratio parameter can be increased by 21%. Compared with the comparison algorithm in the structural similarity parameter and the color information entropy parameter, the proposed algorithm can be improved by 32% and 1.1% respectively.