CAJ | 학술논문

鉴于海雾会造成无人艇视觉系统采集到的视频图像严重降质，进而影响无人艇的目标检测、跟踪以及识别，提出一种融合边缘信息的单尺度 Retinex 快速海雾去除算法。在传统单尺度 Retinex 去雾模型的基础上，结合引导滤波的思想，采用基于边缘信息的高斯滤波来估计亮度分量；再在求得的反射分量中引入原始强度分量来改善反射分量的估计，结合分段灰度变换扩展图像的对比度；最后进行色彩恢复来实现海雾的快速去除。与其他去雾算法进行对比，并从主客观2方面进行图像质量评价。仿真实验结果表明，该算法能有效地减少光晕和噪声，去雾速度快、保边效果好、去雾效果明显。
감우해무회조성무인정시각계통채집도적시빈도상엄중강질，진이영향무인정적목표검측、근종이급식별，제출일충융합변연신식적단척도 Retinex 쾌속해무거제산법。재전통단척도 Retinex 거무모형적기출상，결합인도려파적사상，채용기우변연신식적고사려파래고계량도분량；재재구득적반사분량중인입원시강도분량래개선반사분량적고계，결합분단회도변환확전도상적대비도；최후진행색채회복래실현해무적쾌속거제。여기타거무산법진행대비，병종주객관2방면진행도상질량평개。방진실험결과표명，해산법능유효지감소광훈화조성，거무속도쾌、보변효과호、거무효과명현。
In view of sea fog can lead to the quality decreasing of video image for USV , and affect the detection, tracking, and recognition of targets, a kind of fast single-scale Retinex sea fog removal algorithm fused the edge information was discussed. Firstly, on the basis of traditional single-scale Retinex defogging model, and integrat-ing idea of guided filter, Gaussian filter combined with the edge information was used to estimate the brightness component;then, the original intensity component was lead into the reflection component, so that the estimation of the reflection component was improved so well;and then, combining with the segmented gray scale transform, the contrast of the image was extended;lastly, the color was restored and the rapid sea fog defogging was achieved. Compared with other defogging algorithms from two aspects of subjective and objective, the quality of processed images was evaluated, the simulation experimental results show that proposal algorithm can reduce the halo and noise effectively, defogging is fast and the effect of keeping the edge is good, the effect of defogging is obvious.