何立风, 周广彬, 姚斌, 赵晓, 李笑. 基于引导系数加权和自适应图像增强去雾算法[J]. 微电子学与计算机, 2020, 37(9): 73-77.
引用本文: 何立风, 周广彬, 姚斌, 赵晓, 李笑. 基于引导系数加权和自适应图像增强去雾算法[J]. 微电子学与计算机, 2020, 37(9): 73-77.
HE Li-feng, ZHOU Guang-bin, YAO Bin, ZHAO Xiao, LI Xiao. A haze removal algorithm based on guided coefficient weighted and adaptive image enhancement method[J]. Microelectronics & Computer, 2020, 37(9): 73-77.
Citation: HE Li-feng, ZHOU Guang-bin, YAO Bin, ZHAO Xiao, LI Xiao. A haze removal algorithm based on guided coefficient weighted and adaptive image enhancement method[J]. Microelectronics & Computer, 2020, 37(9): 73-77.

基于引导系数加权和自适应图像增强去雾算法

A haze removal algorithm based on guided coefficient weighted and adaptive image enhancement method

  • 摘要: 针对基于暗通道先验去雾算法易产生图像偏暗、细节信息丢失等现象,提出了一种基于引导系数加权和自适应图像增强去雾算法.首先,对原导向滤波方法进行采样和引导系数加权,快速得到精细化的透射率;然后,利用K-均值聚类将原图像标定为亮色和非亮色区域,约束透射率和大气光值,达到图像噪声抑制和大气光值优化的效果;最后,结合大气散射模型恢复图像,并利用自适应线性对比度增强方法对恢复后的图像进行优化.实验结果表明,与其他代表性去雾方法相比,由本文算法所获得的去雾图像不仅能克服图像失真、细节丢失等问题,同样在主观指标上和客观指标上都能取得较好的结果.

     

    Abstract: Aiming at the dim and detail information loss problem that occurred in dark channel prior dehazing algorithm, a haze removal algorithm based on guided-coefficient-weighted and adaptive image enhancement method was proposed in this paper. Firstly, in order to efficiently refine the transmission, a sampling and guided-coefficient-weighted method was applied to correct the original guided filter algorithm. Then, the K-means clustering method was adopted to mark the bright and non-bright regions according to the original image. It was used to restrict the range of transmission and atmospheric light, thus the noise can be suppression effectively and the atmospheric light can be well optimized. Finally, the image was restored by the atmospheric scattering model and optimized by the adaptive linear contrast enhancement method. Experimental results demonstrated that the proposed method can not only overcome the problem of image distortion and detail information loss, but also more efficient than conventional dehazing methods.

     

/

返回文章
返回