<p>Underwater imagery is often compromised by factors such as color distortion and low contrast, posing challenges for high-level vision tasks. Recent underwater image restoration methods either analyze the input image at full resolution, resulting in spatial richness but contextual weakness, or progressively from high to low resolution, yielding reliable semantic information but reduced spatial accuracy. Here, we propose a lightweight multi-stage network called Lit-Net that focuses on multi-resolution and multi-scale image analysis for restoring underwater images while retaining original resolution during the first stage, refining features in the second, and focusing on reconstruction in the final stage. Our novel encoder block utilizes parallel <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="371_2025_3866_Article_IEq1.gif" Format="GIF" Height="14" Rendition="HTML" Resolution="72" Type="Linedraw" Width="39" /> </InlineMediaObject> <EquationSource Format="TEX">$1\times 1$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1</mn> <mo>×</mo> <mn>1</mn> </mrow> </math></EquationSource> </InlineEquation> convolution layers to capture local information and speed up operations. Further, we incorporate a modified weighted color channel-specific <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="371_2025_3866_Article_IEq2.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="13" /> </InlineMediaObject> <EquationSource Format="TEX">$l_1$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>l</mi> <mn>1</mn> </msub> </math></EquationSource> </InlineEquation> loss (<InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="371_2025_3866_Article_IEq3.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="20" /> </InlineMediaObject> <EquationSource Format="TEX">$cl_1$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>c</mi> <msub> <mi>l</mi> <mn>1</mn> </msub> </mrow> </math></EquationSource> </InlineEquation>) function to recover color and detail information. Extensive experimentations on publicly available datasets suggest our model’s superiority over recent state-of-the-art methods, with significant improvement in qualitative and quantitative measures, such as 29.477 dB PSNR (<InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="371_2025_3866_Article_IEq4.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">$1.92\%$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1.92</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> improvement) and 0.851 SSIM (<InlineEquation ID="IEq5"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="371_2025_3866_Article_IEq5.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="43" /> </InlineMediaObject> <EquationSource Format="TEX">$2.87\%$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>2.87</mn> <mo>%</mo> </mrow> </math></EquationSource> </InlineEquation> improvement) on the EUVP dataset. The contributions of Lit-Net offer a more robust approach to underwater image enhancement and super-resolution, which is of considerable importance for underwater autonomous vehicles and surveillance. The code is available at: <a href="https://github.com/Alik033/Lit-Net">https://github.com/Alik033/Lit-Net</a>.</p>

Harnessing multi-resolution and multi-scale attention for underwater image restoration

期刊论文
发表日期：

2025年3月19日

查看全文

Alik Pramanick,
Arijit Sur,
V. Vijaya Saradhi

摘要

Underwater imagery is often compromised by factors such as color distortion and low contrast, posing challenges for high-level vision tasks. Recent underwater image restoration methods either analyze the input image at full resolution, resulting in spatial richness but contextual weakness, or progressively from high to low resolution, yielding reliable semantic information but reduced spatial accuracy. Here, we propose a lightweight multi-stage network called Lit-Net that focuses on multi-resolution and multi-scale image analysis for restoring underwater images while retaining original resolution during the first stage, refining features in the second, and focusing on reconstruction in the final stage. Our novel encoder block utilizes parallel $1\times 1$ $1 \times 1$ convolution layers to capture local information and speed up operations. Further, we incorporate a modified weighted color channel-specific $l_1$ $l_{1}$ loss ( $cl_1$ $c l_{1}$ ) function to recover color and detail information. Extensive experimentations on publicly available datasets suggest our model’s superiority over recent state-of-the-art methods, with significant improvement in qualitative and quantitative measures, such as 29.477 dB PSNR ( $1.92\%$ $1.92 %$ improvement) and 0.851 SSIM ( $2.87\%$ $2.87 %$ improvement) on the EUVP dataset. The contributions of Lit-Net offer a more robust approach to underwater image enhancement and super-resolution, which is of considerable importance for underwater autonomous vehicles and surveillance. The code is available at: https://github.com/Alik033/Lit-Net.