Unveiling the depths of underwater image enhancement with spatial blended CNN: Diving into clarity

Abstract

This study introduces a pioneering approach called spatial blended convolutional neural network (SBCNN) tailored specifically for enhancing the visual quality of underwater images. Traditional methods often struggle with the challenges posed by underwater environments, such as light absorption, scattering, and colour distortion. SBCNN addresses these challenges by combining spatial techniques with convolutional neural networks (CNNs), leveraging the strengths of both approaches. The architecture of SBCNN is designed to effectively capture spatial details and structural information inherent in underwater images. It incorporates spatial CNN branches, which are specialized components dedicated to analyze and enhance spatial features within the image. By integrating these spatial branches with traditional CNN layers, SBCNN can effectively address the unique characteristics of underwater imagery. To assess the performance of SBCNN, the study conducted comprehensive experimental evaluations using three different datasets: UIEB, EUVP, and UFO-120 and quantitative metrics such as peak signal-to-noise ratio (PSNR), mean square error (MSE), and Structural Similarity Index (SSIM) were utilized to compare the results with existing methods. The findings from these evaluations demonstrated significant improvements achieved by SBCNN over traditional techniques, indicating its effectiveness in enhancing the visual quality of underwater images. Furthermore, to validate the generalizability of the proposed method, cross-dataset testing was conducted on the EUVP ImageNet dataset and UIEB dataset.