DWFUIR: deep weighted least square filter for underwater image restoration

Underwater image processing has been a major field of study in ocean exploration, and numerous convolutional neural network based techniques for underwater picture improvement and restoration have been developed over time. Real-world underwater photos typically have a variety of quality issues, including colour casts, poor contrast, decreased visibility, and so on, because of attenuation and scattering of beam in water medium. Thus, in real-world applications, these quality flaws have a negative impact on underwater photographs. To solve these difficulties, in this paper, we suggest an efficient Deep Weighted Least Square Filter for underwater image restoration (DWFUIR). A novel building block for Deep Convolutional Neural Network (DCNN) acquires an end-to-end mapping among small resolution inputs and produce better quality outputs. Such a layer contains learnable parameters, which can be integrated with Weighted Least Square Filter and jointly optimized through training. By integrating suggested layer with DCNNs, DWFUIR can generate better restored, edge-preserving outputs. Experimental evaluations demonstrate that DWFUIR performs better than current restoration techniques, obtaining higher scores in quantitative and qualitative evaluations while successfully maintaining edges. These findings show that the suggested method offers a reliable and effective way to restore underwater images, with a great deal of promise for real-world uses in marine exploration and research.