结合最小颜色通道图与传播滤波的单幅图像去雾算法研究

发布时间：2018-10-24 22:58

【摘要】：针对单幅图像去雾中边缘区域去雾不彻底及黑斑现象,提出了一种结合最小颜色通道图与传播滤波的图像去雾算法。该方法首先基于双区域滤波实现大气透射率的初始估计,然后引入最小颜色通道图作为参考图像,采用传播滤波优化初始透射率,优化后的透射图因与参考图像具有相似的边缘特性,能有效避免图像景深突变边缘像素点透射率估值出现偏差的问题,同时去除了初始透射图中冗余的纹理信息,最后利用L-BFGS自适应恢复大气光矢量,基于大气散射模型复原无雾图像。实验结果表明,所提算法对图像景深突变边缘具有更为精确的透射率估算值,能有效保持边缘和细节,图像景深均匀区域具有较好的空间平滑特性,复原后的图像具有较高清晰度和丰富色彩度。
[Abstract]:Aiming at the incomplete defogging and black spot phenomenon in the edge region of a single image, an image de-fogging algorithm combining the minimum color channel diagram and the propagation filter is proposed. The method first realizes the initial estimation of atmospheric transmittance based on two-region filtering, then introduces the minimum color channel diagram as the reference image, and uses the propagation filter to optimize the initial transmittance. The optimized transmission image has similar edge characteristics to the reference image, which can effectively avoid the problem of error in the estimation of the transmittance of the edge pixel of the depth of field mutation. At the same time, the redundant texture information in the initial transmission image is removed. Finally, the atmospheric light vector is self-adaptively restored by L-BFGS, and the fog free image is reconstructed based on the atmospheric scattering model. The experimental results show that the proposed algorithm has a more accurate transmittance estimation for the abrupt edge of the depth of field, and can effectively preserve the edges and details, and the uniform region of the depth of field has a better spatial smoothing property. The reconstructed image has high definition and rich color.
【作者单位】：湖南大学电气与信息工程学院;湘潭大学信息工程学院;湘潭大学控制工程研究所;湖南大学信息科学与工程学院;
【基金】：国家自然科学基金资助项目(No.61573299,No.61673162,No.61672216,No.61602397) 湖南省教育厅基金资助项目(No.15C1328)~~
【分类号】：TP391.41

【相似文献】