A hybrid method for underwater image correction.
Pattern Recognition Letters(2017)
摘要
We estimate transmission by combining learning-based strategy with optical property.A global background light estimation algorithm specialized for underwater images is proposed.An underwater image dataset which includes 45 underwater images is collected. Underwater images surfer from serious color deviation and blurring due to the effects of light absorption and scattering. In this letter, a hybrid method, which includes color correction and underwater image dehazing, is proposed to improve the visual quality of degraded underwater images. Firstly, an efficient color correction algorithm is applied to remove color casts of underwater images. Then, underwater image dehazing method is proposed to improve the visibility of underwater images, which includes a global background light estimation algorithm specialized for underwater images and a medium transmission estimation algorithm based on the combination a regression model with the characteristics of light traveling in water medium. Since there is no available dataset in this relatively new research area, a dataset which includes 45 underwater images with a wide variety of contents is collected. Subjective and objective performance evaluations demonstrate that the proposed method significantly improves both color and visibility of degraded underwater images, and is comparable to and even outperforms several state-of-the-art methods.
更多查看译文
关键词
Underwater image correction,Color casts,Image degradation,Visibility recovery
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络