Underwater Images Enhancement by Revised Underwater Images Formation Model

Pei, Soo‐Chang; Chen, Chia-Yi

doi:10.1109/access.2022.3213340

Cited by 18 publications

(1 citation statement)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Author used Structural Similarity metric also called as SSIM to compare the results and the method proposed in this paper has achieved the best results with SSIM 85%. [3] Pei, S. C., & Chen, C. Y worked on the problem of color distortion, low contrast and loss of the human activity due to light scattering and absorption. Author proposed a revised underwater dehazing model aiming to eliminate the color of water directly while solving the problem of haze in the underwater images.…”

Section: Literature Surveymentioning

confidence: 99%

Image Enhancement of Underwater Images using Deep Learning Techniques

Sravani¹,

Sai²,

Noomika³

et al. 2023

Int. J. Res. Publ. Rev.

View full text Add to dashboard Cite

The quality of underwater images is very important for research purpose and ocean exploration. The market for high-quality underwater images is expanding exponentially. Despite this, there are a number of issues including blurriness, dim lighting, poor contrast and color fidelity. Light attenuation, absorption and scattering lower the quality. As a result, it is essential to enhance underwater images, and different image processing techniques are used for enhancement .The deep learning methodology used in this study improves image quality and produce the best outcomes when compared to conventional image enhancing techniques. The architecture used is GAN with the combination of Image processing algorithms as a preprocessing step for implementing deep learning models .The algorithms include Histogram Equalization and Dark Channel Prior. These methods normalizes the contrast of images and removes the hazing or dark parts of the images. The dataset used in this study is EUVP (Paired) which contains about 2186 real-world underwater images. Three parameters, UICM, MSE and PSNR are used to assess quality of enhanced images. The optimum technique for improving the images is determined by comparing the performance of the models after the dataset has been trained through them.

show abstract

Section: Literature Surveymentioning

confidence: 99%