Image Enhancement of Degraded Sand-dust Images Based on Channel Compensation and Brightness Partitioning
Pingping Xiang, Chunmei Chen, Guihua Liu, Zhongxiang Pang, Jiashuo Zhang, Junlei Hu · 2023
This paper proposes a new method for enhancing sand-dust images by correcting color deviation and improving contrast, which is aimed at addressing the visual interference caused by sand-blind conditions for pilots. Firstly, to compensate for the color attenuation of the blue channel and better correct the yellowish appearance of sand-dust images, a channel compensation module based on the standard deviation of the yellow histogram is proposed. Secondly, a color normalization module that maintains the yellow mean value is designed to correct the color imbalance issue by making the red, blue, and green histograms similar. Finally, a partition-based enhancement module based on the brightness feature is constructed to decompose the brightness space V into a highlight region B and a dark region D, which are enhanced by limiting contrast adaptive histogram equalization and nonlinear transformation, respectively. The experimental results demonstrate that the proposed algorithm achieves good enhancement performance on sand-dust images with different degrees of color deviation, and outperforms SDIE in terms of NIQE, NBIQA, BRISQUE, and NPQI metrics, achieving performance improvements of 14.22%, 19.87%, 13.40%, and 8.96%, respectively, with more effective significance detection performance.