Human Visual System Based-Quality Evaluation Model for Underwater Image

Underwater images typically have feature issues including blur, low contrast, and imbalanced colours owing to the absorption and scattering effects of water. Existing raw image quality evaluation method cannot be easily applied to underwater images as it does not consider the particularity of underwater imaging. This paper proposes a new unreferenced underwater image quality evaluation method closely related to subjective perception. The chromaticity feature (Col), the contrast feature (Con) based on visual cortex of human brain, and the sharpness feature (Sharp) reflecting the richness of image information constitute the underwater image quality evaluation model, which is referred to as CCS. The features are sensitive to the physical properties of water, and the Human Visual System (HVS) is susceptible to changes in visual properties such as colour, contrast, and edge structure. A self-built small underwater image is used to verify the proposed method's performance. Experiments conducted on the dataset with CPDB, BRISQUE, UCIQE, and UIQM as non-reference evaluation algorithms reveal that the CCS algorithm is highly in line with human visual perception and can accurately evaluate the underwater images.