Single Underwater Image Restoration Using Adaptive Attenuation-Curve Prior

Yi Wang, Hui Liu, Lap Pui Chau

Research output: Journal article publicationJournal articleAcademic researchpeer-review

71 Citations (Scopus)

Abstract

Underwater imaging is an important topic in maritime research. Due to the existence of dust-like particles in water medium, underwater images are vulnerable to the effect of low contrast and color cast. In this paper, we propose a novel underwater image restoration method based on a non-local prior, namely, adaptive attenuation-curve prior. This prior relies on the statistical distribution of pixel values. That is, all pixel values of a clear image can be partitioned into several hundred distinct clusters in RGB space, and the pixel values in each cluster will be distributed on a curve with a power function form after attenuated by water in varying degrees. Specifically, we can estimate the transmission for each pixel according to its distribution on the curves. Then, we estimate the attenuation factor to compensate for the transmission. To prevent over saturation and reduce the noise of the recovered images, we propose the saturation constraints to adjust the transmission of the three color channels. Qualitative and quantitative results demonstrate that our proposed method can achieve better performance, compared with the state-of-the-art approaches. Moreover, our proposed method can be further extended to restore other kinds of degraded images, such as hazy images.

Original languageEnglish
Pages (from-to)992-1002
Number of pages11
JournalIEEE Transactions on Circuits and Systems I: Regular Papers
Volume65
Issue number3
DOIs
Publication statusPublished - Mar 2018
Externally publishedYes

Keywords

  • attenuation-curve prior
  • image restoration
  • light attenuation
  • transmission estimation
  • Underwater image
  • waterlight

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Single Underwater Image Restoration Using Adaptive Attenuation-Curve Prior'. Together they form a unique fingerprint.

Cite this