摘要
为解决传统暗通道不适用于大面积天空区域,容易造成去雾图像失真的问题,提出一种结合暗亮通道先验的远近景融合去雾算法。首先,利用改进的二维Otsu图像分割算法,混合近景和远景区域的暗通道,并基于最优的客观质量评价指标对近景和远景区域设置混合暗通道的自适应调节参数;其次,针对真实物理场景中大气光并非均匀不变常量的问题,建立暗亮通道融合模型,并计算大气光图;为了提升处理速度,在不降低恢复质量的前提下,选取与原图对应的灰度图作为引导图像对透射率图进行细化;最后,采用基于视觉感知的亮度/颜色补偿模型对图像修正,提高了复原图像的对比度和色彩饱和度。实验结果表明,所提算法在主观和客观角度均取得最好的效果,其中客观指标PSNR在数值上比He的算法平均高出24.04%。由此得出,通过所提算法复原的图像更加清晰、细节信息和结构更加明显,更适于人眼的观察,验证了算法的有效性。
A far and near scene fusion defogging algorithm based on the prior of dark-light channel is proposed to solve the problem that traditional dark channel is not suitable for large sky area and it is easy to cause the distortion of dehazed image.Firstly,an improved two-dimensional Otsu image segmentation algorithm is utilized to mix the dark channels in the close and distant areas,and the adaptive adjustment parameters of the mixed dark channels are calculated based on the optimal objective quality evaluation index for the close and distant areas.Secondly,aiming at the problem that atmospheric light is not uniform and constant in real physical scenes,a dark-light channel fusion model is established to calculate the atmospheric light map.Furthermore,in order to improve processing speed,the grayscale image corresponding to the original image is selected as a guide image to refine the transmittance image without reducing restoration quality.Finally,the brightness/colour compensation model based on visual perception is used for image correction to improve the contrast and colour saturation of the restored image.Experimental results show that the proposed algorithm achieves the best results from both subjective and objective perspectives,in which the objective index PSNR is 24.04%higher than that of He’s algorithm on average.It is concluded that the image recovered by the proposed algorithm is clearer,with more obvious details and structure,and is more suitable for human eyes to observe,which verifies the effectiveness of the algorithm.
作者
高涛
刘梦尼
陈婷
王松涛
蒋硕
GAO Tao;LIU Mengni;CHEN Ting;WANG Songtao;JIANG Shuo(School of Information Engineering, Chang’an University, Xi’an 710000, China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2021年第10期78-86,共9页
Journal of Xi'an Jiaotong University
基金
国家重点研发计划资助项目(2019YFE0108300)
国家自然科学基金资助项目(62001058)
陕西省重点研发计划资助项目(2019GY-039)。
关键词
图像去雾
图像分割
混合暗通道
亮通道
图像修正
image defogging
image segmentation
mixed dark channel
light channel
image correction
