Single image dehazing enhancement based on retinal mechanism

• Published in: Multimedia tools and applications Vol. 83; no. 21; pp. 61083 - 61101
• Main Authors: Lei, Lei, Cai, Zhe-Fei, Fan, Ying-Le
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2024; Springer Nature B.V
• Subjects: Color; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Datasets; Image contrast; Image enhancement; Image quality; Luminous intensity; Multimedia Information Systems; Retina; Signal to noise ratio; Special Purpose and Application-Based Systems; Synthetic data; Track 6: Computer Vision for Multimedia Applications; Single-opponent receptive field; Single image dehazing; Crossover inhibition; Gap junction coupling
• ISSN: 1573-7721; 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-17935-w

Based on the hierarchical transmission and interaction response characteristics of visual information in retina, we propose a single image dehazing enhancement computational model that simulates the multiple level physiological mechanisms of retina. Firstly, according to the characteristics of gap junction between horizontal cells followed by the dynamic changing properties of light intensity, an adaptive adjustment model of receptive fields based on the luminance information is constructed to protect the detail information at the edge of hazy images. Secondly, simulating the crossover inhibition mechanism of ON and OFF pathways, the AII amacrine cell network model is constructed to inhibit the OFF pathway through the ON pathway, thus expanding the dynamic range of the hazy images. Finally, according to the dynamic correlation between the receptive field characteristics of ganglion cell and local contrast, a single-opponent receptive field (SORF) dynamic adjustment model based on local contrast information is constructed to enhance the contrast of hazy images. Natural image dataset and the composite image dataset RESIDE-OTS are used as the experimental subjects. In the natural image dataset, Fog Aware Density Evaluator (FADE) ranked 3rd in the average score and 1st in the Natural Image Quality Evaluator (NIQE) average score; Compared with the highest index in other methods, the peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) in the synthetic image dataset increased by 7.20% and 4.16%, respectively. Experiments demonstrate that our method enhances the details and contrast while retaining the original color characteristics of the image, and improves the problems of color distortion and halation, which provides a new idea for the internal mechanism and application of brain vision.