Underwater Polarization Imaging Recovery Based on Polarimetric Residual Dense Network

Abstract

Application of deep-learning to polarization imaging technology for image restoration has led to many technological breakthroughs, especially in underwater image recovery and recognition. In this work, a four-input deep learning model with the Polarimetric Residual Dense Network is proposed for underwater image recovery. The diverse polarization component images are trained and tested in different processes in the network for the recognition and dehazing by considering the physical model of polarization dehazing imaging. Our study reveals that the proposed method can efficiently recover the hazed images, and provide good performance for improving the quality of image restoration even in a high-turbidity complex underwater environment.

A fusion adaptive recognition network based on intensity and polarization imaging

“…By obtaining the degree of polarization and polarization angle as well as adopting certain computing methods, the subtle depth information of the far-field target can be deciphered, and the three-dimensional polarization imaging of the target can be realized 6 . With the improvement of hardware arithmetic power and the rapid development of deep learning, the combination of polarization and artificial intelligence has been gradually emphasized and achieved good results.Zhang et al achieved polarization fusion at the feature level and decision level of the input data by combining dual polarization features with geometric feature images, and achieved good results in synthetic aperture radar ship identification 7 .Wang et al used the MBINEDN network to extract the feature information of polarization image and infrared image respectively, and realized the effect of semantic segmentation of vehicle and road environment which is better than single-mode features in low visibility environment 8 .Xiang et al used laser beam as the active light source for polarization imaging, and trained the obtained information of linearly polarized images with different angles and circularly polarized images as inputs to the residual dense network, and realized the effect of highly turbid image restoration in complex underwater environments 9 . In the field of geology, Zhang et al effectively improved the recognition accuracy of rock sheet images under microscope by fusing orthogonally polarized images and single polarized images of rock sheets from different levels to enhance the dimensions of information as extracted features 10 .…”

Polarization-Enhanced Underwater Detection Method for Multiple Material Targets Based on Deep-Learning

“…UE to complex scattering, low visibility, and cluttered targets, target detection in underwater environments poses significant challenges [1]. Traditional imaging techniques often fail to capture the details and surface characteristics of underwater targets, making it difficult to detect and recognize the material features of targets [2].…”