Underwater Object Segmentation Integrating Transmission and Saliency Features

Abstract

Various types of knowledge and features have been explored for level set-based segmentation. On the ground, the prior knowledge and carefully-designed features perform well to identify the foregroundbackground contrast, which improves the performance of the segmentation method for complicated and distorted data. However, this is not the case for underwater environments, since the features available on the ground are not suitable for challenging underwater environments. Thus, underwater image segmentation currently lags behind ground-based segmentation. In this paper, novel cues and a suitable model formulation for object segmentation from underwater images are proposed. We consider the special haze effect over underwater images and extract an informative feature (transmission feature) from haze condensation. The saliency feature is also used for underwater object segmentation. Consequently, in our method, the objectbackground difference can be presented by these features on two levels, i.e., the edge-level transmission and region-level saliency features. These two types of features are integrated into a unified level set formulation to propose a solution that handles the challenging issues in underwater object segmentation. The experimental comparisons of our method with other methods comprehensively demonstrate the satisfactory performance of our method.

The Use of Saliency in Underwater Computer Vision: A Review

“…On the other hand, it suggests that saliency, although being a somewhat subjective notion, is a powerful concept that has still much to say in underwater vision. [21] 2018 Object detection and CNN-based classification Itti Atallah et al [23] 2005 Object detection Entropy-based Wang et al [24] 2013 Detection & Segmentation Itti Chen et al [27] 2014 Object detection Spectral residual Chuang et al [48] 2016 Initialization of object recognition Phase Fourier Transform Zhu et al [34] 2017 Detection & Segmentation Saliency map based on contrast, position, and correspondence Sanchrez-Torres et al [99] 2018 Segmentation Ad hoc based on morphological operators Huo et al [26] 2018 Detection & 3D Reconstruction Aggregation of salient superpixels Kumar et al [36] 2019 Shape reconstruction using edge-based active contours Itti Chen et al [40] 2019 Segmentation using region-based active contours HFT Barat et al [35] 2010 Segmentation using active contours featuring saliency in initialization Itti Kumar et al [31] 2019 Moving object detection Multiple frames difference Zhu et al [50] 2019 Template Matching Spectral residual Jian et al [51] 2018 Object detection QDWB Jian et al [53] 2018 Object detection QDWB + PD + LC Johnson-Roberson et al [47] 2010 Classification Entropy-based Cong et al [32] 2019 Saliency-based Object Detection Saliency map obtained by Deep Convolutional Neural Network Harrison et al [56] 2011 Texture segmentation Co-occurence matrices and ensemble of distance…”

“…Initialization seeds for the active contours are obtained using the method reported in Reference [29], which is also discussed in Section 3.2 of the present paper. Another approach to object segmentation based on active contours is presented in Reference [40], where saliency is used as a data-driven term to favor convergence of the contours to the boundaries of the desired object of interest. The authors employ a formulation of active contours, based on the so-called level-set approach, in which the contours to be sought are represented as the zero set of a real function φ having the image as its domain.…”

DPANet: Dual Pooling‐aggregated Attention Network for fish segmentation

“…Wang et al [54] proposed an effective fuzzy Cmeans method based on particle swarm optimisation specifically for underwater optical segmentation. By modelling the effect of haze in underwater images, Chen et al [53] proposed a new method for underwater object segmentation where edge-level transmission features and region-level significant cues were blended together and computed using a level set formulation. Zhang et al [52], on the other hand, combined the fractal theory with the characteristics of underwater images and proposed a Brownian random field method that can effectively help in the segmentation of underwater images.…”

Dual-fusion Active Contour Model with Semantic Information for Saliency Target Extraction of Underwater Images

“…[12] and Ref. [17], which also introduced saliency images as semantic information. Since we cannot get the source codes of Ref.…”

“…However, this method required that the target region of the underwater image has uniform grayscale. Chen et al [17] integrated the transmission map and the saliency map into a unified level set formulation to extract the salient target contours of the underwater images.…”