Underwater Image Quality Assessment and Enhancement Using Active Inference

The global attention toward the marine environment has been increasing due to the abundance of debris found in shallow and open seas, coasts, and even the seabed. The use of submersibles with debris detection systems can help in surveying and collecting the same. However, images obtained for this purpose are plagued with numerous issues in the field of image processing, such as heavy light distortion and scattering, and the disappearance of red hue as we descend into deeper waters. To address this, the existing NRIQA-GAN model with Active Inference constraints is used to evaluate the value of underwater pictures, and UGAN-P model is employed to enhance the images. Image segmentation has also been performed as a preliminary step toward future work on deep-sea debris detection using object detection methods.