基于修正散射模型的水下图像复原
首发时间:2019-08-27
摘要:由于光在水下的传输衰减特性,水下图像普遍存在对比度低、模糊和颜色失真等问题。目前的研究大多集中于背景光及介质透射率估计上,而本文重点研究了水下成像模型,从海洋光学原理角度出发,针对经典模型的局限性,提出一种修正的水下成像模型。首先,深入分析光在水下的吸收衰减特性,在简化大气散射模型的基础上,将水体背景光融入到模型的直接衰减项;其次,考虑到水下红光迅速衰减的问题,采用红通道的逆通道进行补偿;然后,使用基于四叉树的分级搜索算法估计水体背景光值;最后,在修正的成像模型基础上,结合水下暗通道先验估计介质透射率进而复原水下图像。实验结果表明,本文算法复原后的水下图像色彩自然,能有效恢复出远景区域的细节信息,图像对比度、色度和饱和度的综合评价指标整体优于对比算法,且模型复杂度低,适用于不同类型的水下退化图像。
关键词: 图像处理 水下图像复原 成像模型 红通道补偿 暗通道先验
For information in English, please click here
Underwater Image Restoration Based on Modified Scattering Model
Abstract:Due to the attenuation characteristics of light transmission, there are many problems in underwater image, such as low contrast, blurring and color distortion. Most of the current research focus on the estimation of background light and transmission. This paper focuses on the underwater imaging model. In view of the limitations of the classical model, a modified underwater imaging model is proposed from the perspective of ocean optics. Firstly, the absorption attenuation characteristics of light are analyzed in depth. Based on simplified atmospheric scattering model, the background light of water body is integrated into the direct attenuation term of the model. Secondly, considering the rapid attenuation of underwater red light, the inverse channel of the red channel is used to compensate the attenuation. Then, the background light value of water body is estimated by the hierarchical search algorithm based on quadtree. On the basis of the imaging model, the underwater image is restored by estimating the transmittance combined with underwater dark channel Prior. The experimental results show that the restored underwater image has the natural color and can effectively restore the detail information. The comprehensive evaluation index of image contrast, color and saturation are better than the contrast algorithm as a whole, and the model complexity is low. It is suitable for different types of underwater degraded images.
Keywords: image processing underwater image restoration imaging model red channel compensation dark channel prior
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于修正散射模型的水下图像复原
评论
全部评论0/1000