We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.IV

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Image and Video Processing

Title: Underwater Image Enhancement based on Deep Learning and Image Formation Model

Abstract: Underwater robots play an important role in oceanic geological exploration, resource exploitation, ecological research, and other fields. However, the visual perception of underwater robots is affected by various environmental factors. The main challenge now is that images captured by underwater robots are color-distorted. The hue of underwater images tends to be close to green and blue. In addition, the contrast is low and the details are fuzzy. In this paper, a new underwater image enhancement algorithm based on deep learning and image formation model is proposed. Experimental results show that the advantages of the proposed method are that it eliminates the influence of underwater environmental factors, enriches the color, enhances details, achieves higher scores in PSNR and SSIM metrics, and helps feature key-point point matching get better results. Another significant advantage is that its computation speed is much faster than other methods.
Subjects: Image and Video Processing (eess.IV)
Cite as: arXiv:2101.00991 [eess.IV]
  (or arXiv:2101.00991v2 [eess.IV] for this version)

Submission history

From: Xuelei Chen [view email]
[v1] Mon, 4 Jan 2021 14:03:50 GMT (2415kb)
[v2] Thu, 7 Jan 2021 14:45:42 GMT (4419kb)

Link back to: arXiv, form interface, contact.