Underwater Single Image Dehazing Using the Color Space Dimensionality Reduction Prior

Abstract

Underwater images suffer from low visibility and contrast caused by absorption and scattering, which leads to haze and some further limitations. The existing underwater single image dehazing methods cannot achieve a balance between the performance and computational complexity, and are difficult to produce satisfactory results in the regions with large distance. To overcome these problems, we propose a new underwater single image dehazing method, which includes an improved background light estimation based on the quad-tree subdivision iteration algorithm, and a novel transmission estimation method. For the background light estimation, we introduce a robust score for each region of the image, which can evaluate the region from both smoothness and color. For the transmission estimation, we propose the color space dimensionality reduction prior (CSDRP), which allows conversing an image from the three-dimensional RGB color space to a 2D color space, namely the UV color space. In the UV color space, by clustering the pixels into mounts of haze-lines and carefully setting the haze-free boundary, the transmission map can be figured out and used to produce an excellent dehazed image. Experimental results show that our method has competitive effects compared with mainstream underwater single image dehazing methods. INDEX TERMS Underwater image dehazing, contrast enhancement, image enhancement, scattering removal.

Underwater Images Enhancement by Revised Underwater Images Formation Model

“…In this section, we will evaluate the performance of the proposed algorithm with a comparison of quantity and quality. For the general algorithms, we compare our results with the Haze-lines methods [14], the red channel method [9], the fusion method [50], dark channel methods WGDCP [10], Color Space Dimensionality Reduction Prior [25] and sea-thru [13]. For the deep learning-based method, we selected several effective algorithms, including the GAN-based method, i.e., Water-GAN [17] and FUnIE-GAN [52], and Begin Contrast Enhancement 1: Contrast Enhancement using Eq.…”

“…Figure 14 compared the results with [25], [13], [14], and WGDCP [10]. We see the red boxes and their details in Fig.…”

“…14, and the red boxes show that our method performs a clearer scene than the others. [25] recovered the color well; however, they make the scene over-exposure. Although [9] shows the structures of the recovered images, they cannot recover the vivid color.…”

“…PCQI, PCQI UIQM and UCIQE comparison on USR-248 underwater dataset USR-248 Underwater Dataset Method Metric Original Image Ancuti et al [50] Berman et al [14]Xie et al's method[9] Sea Thru[13] WGDCP[10] Liu et al[25] Water-GAN[17] FUnIE-GAN[52] UIE-WD[53] …”

Color Correction and Local Contrast Enhancement for Underwater Image Enhancement

“…For the third class of methods using prior knowledge [19]- [23]. The dark channel prior (DCP) proposed by Heet al [19] is used for outdoor image dehazing.…”