Polarization image defogging method and device based on non-supervision weight depth model

The invention discloses a polarization image defogging method and device based on a non-supervision weight depth model, and relates to the field of image defogging processing, and the method comprises the steps: taking a deentanglement representation learning thought as a starting point, directly solving the polarization information of light waves according to a stokes theory, directly inputting a first image and a second image into the non-supervision weight depth model, and carrying out the defogging of a polarization image. A dehazed polarization image is output through a coding layer, a fusion layer and a decoding layer in sequence, in a model training method, two weights are designed in a weight measurement layer, so that retention modes of different modal information are considered, the two weights are counted into a loss function to train a non-supervision weight depth model, and a non-supervision weight depth model is obtained. And finally, long-distance scene defogging is realized. A focal plane type