Focal plane polarization image denoising method based on deep neural network

Abstract

The invention discloses a focal plane polarization image denoising method based on a deep neural network. The method comprises the following steps: creating a polarization image data set; designing adeep neural network PDRDN, wherein the network learns a residual image between the noise image and the noiseless image; optimizing the network parameters by using a customized loss function until themodel parameters are stable; and inputting an image to be denoised into the trained network to obtain an output residual image, and then predicting a clean image by using the noise image and the residual image. The method can achieve a better denoising effect, is used for improving the quality of a polarization image, and reduces the impact on polarization information from noise.

Description

technical field [0001] The invention relates to the technical field of polarization imaging detection, in particular to a method for denoising a focal plane polarization image based on a deep neural network. Background technique [0002] Polarization imaging has a wide range of applications in many fields. The sub-focal plane camera can collect images of four polarization directions at the same time, and it is more and more widely used in various fields. However, in practical applications, it is inevitable that the collected polarization images have noise interference, which affects the reconstruction of polarization information. accuracy. Therefore, the denoising technology for polarization imaging of split focal plane cameras is of great significance. The polarization image collected by the focal plane camera records the information of the four polarization directions, and has physical correlation. Although the simple replication of the traditional denoising method is fe...

AI Technical Summary

Problems solved by technology

The polarization image collected by the focal plane camera records the information of the four polarization directions, and has physical correlation. Although the simple replication of the traditional denoising method is feasible, the reconstruction effect of the polarization information is limited.

The current denoising method for polarized images mainly has two shortcomings: First, it is simply assumed that the noise type is Gaussian, but the real environmental noise distribution is more complex, and only simulating the real noise through Gaussian noise will cause the algorithm to fail in the real noise scene. restricted

Second, the traditional polarization image denoising needs to rely on a lot of prior knowledge to determine the algorithm parameters, it is difficult to achieve the optimal solution of the algorithm

Images