An underwater-imaging-model-inspired no-reference quality metric for images in multi-colored environments. (1st April 2022)

• Record Type: Journal Article
• Authors: Luo, Zhihang; Tang, Zhijie; Jiang, Lizhou; Wang, Chi
• Is Part Of: Expert systems with applications. Volume 191(2022)
• Publication Date: 2022-04-01
• Subjects: No-reference quality metric -- Underwater images -- Multi-colored environments -- Real-time image processing; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2021.116361
• Languages: English
• ISSNs: 0957-4174

Abstract

Graphical abstract: Highlights: The proposed metric is applied to multi-colored underwater environments. The metric obtains a higher correlation between the predictions and true quality. The calculation of true quality scores is based on the measured turbidity. The proposed metric has the characteristics of strong generalization ability. The proposed metric performs well in real-time image processing. Abstract: The color of an underwater target is different under different lighting conditions and underwater environments. To date, no quality metric has been proposed for images of underwater targets in multi-colored environments. In this paper, we proposed a no-reference quality metric for images of underwater targets in multi-colored environments (QMICE). This metric is a weighted combination of colorfulness index, contrast index and sharpness index. The colorfulness index is used to measure the color loss caused by absorption. The contrast index and sharpness index are used to measure the blurring caused by scattering. The weighted coefficients of the three indexes are calculated by multiple linear regression (MLR). For the contrast index and sharpness index, we proposed a grayscale conversion method that can adaptively adjust the coefficients of red, green, and blue (RGB) values to enhance their generalization ability under multi-colored environments. During the calculation of weighted coefficients, the quality scores based on turbidity are regarded as the true quality scores. It is more reliable than subjective assessment scores. The experimental results show that compared with the leading underwater image quality metrics available in the literature, the proposed metric has the best correlation between the metric predictions and the true quality scores. More importantly, QMICE can also be used to process underwater images in real time and evaluate the performance of underwater image restoration algorithms.