Current Issue Cover
混合先验与加权引导滤波的图像去雾算法

李喆,李建增,胡永江,张岩()

摘 要
目的 图像去雾是降低雾、霾、沙等低能见度成像环境对图像的退化影响,提高图像信息获取质量的过程。为了消除先验盲区,同时进一步提高去雾图像边缘细节的清晰度,提出一种混合先验与加权引导滤波的图像去雾算法。方法 首先改进大气光值估计方法,提高大气光值估计的准确性。然后利用混合先验理论求取双约束区域的大气透射率,一定程度上消除了先验盲区,提高了去雾算法的鲁棒性。最后利用加权引导滤波算法优化透射率图,提高了图像边缘细节的清晰度。结果 本文以通用去雾测试图像和小型无人机拍摄的雾天图像作为实验对象,通过对比分析4种组合步骤算法的复原效果,验证本文各步骤改进方法的合理性与整体算法的优越性。实验结果表明:混合先验理论改善了暗原色先验在明亮区域的失真现象和颜色衰减先验对浓雾处理上的不足,取得了较好的视觉效果;加权引导滤波改善了图像边缘模糊的现象,使复原后的图像边缘细节更加清晰;相较传统算法,本文算法视觉效果更好,去雾图像边缘细节更加明显,综合评价指标均值提升幅度较大。结论 针对有雾图像复原,通过理论分析和实验验证,说明了本文改进的各步骤具有一定的优越性,所提的算法具有较强的鲁棒性。
关键词
Mixed Prior and Weighted Guided Filter Image Dehazing algorithm

李 喆,李 建增,胡 永江,张 岩(Department of UAV engineering, Army Engineering University, Shijiazhuang 050003, China)

Abstract
Objective Image dehazing is a process of reducing the degradation effect of low-visibility imaging environment such as fog, sputum and sand, and improving the quality of image information acquisition. It mainly solves the problems of image feature information blur, low contrast, gray level concentration, color distortion and the like. At present, image dehazing methods are mainly divided into two categories: image restoration and image enhancement. By analyzing the degradation mechanism of the image and using prior knowledge or assumptions, the dehazing algorithm based on image restoration is used to establish a physical model of image degradation to restore the clear image in a targeted manner. Compared to image enhancement algorithms, they are more targeted. Deblurring is better and image information is more complete. Therefore, it is of great significance to its research. In order to eliminate the prior blind zone and improve the sharpness of the edge detail of the haze-free image, an image dehazing algorithm based on mixed prior and weighted guided filter (MPWGF) is proposed.Method Firstly, in order to reduce the limitation of atmospheric light value estimation and make full use of the advantage of mixed prior conditions, a new method of atmospheric light value estimation is proposed. The pixel positions of 0.1% before brightness in dark channel map and depth map are extracted respectively, and the coordinate points extracted from the two images are compared. If there are two images at the same time, the coordinate points are retained, otherwise, the values with the highest brightness corresponding to the remaining coordinate points in the original image are eliminated. This method can eliminate outliers to some extent and improve the accuracy of atmospheric light estimation. Then, the mixed prior theory is used to calculate the atmospheric transmission of the double constraint region, which eliminates the prior blind zone to a certain extent. This improves the robustness of the dehazing algorithm. Both dark channel prior (DCP) and color attenuation prior (CAP) have better recovery effects, and can better compensate for the existence of the prior blind zone. Therefore, an effective region segmentation method is proposed to segment the bright and foggy regions of blurred images. According to the regional characteristics, using DCP and CAP to get atmospheric transmittance maps to solve the problem that the prior blind region which exists in a single prior estimation method affects the robustness of the restoration algorithm. Finally, the adaptive guided filter algorithm is used to optimize the transmission map, which further improves the sharpness of the image edge details. In order to eliminate the halo and block artifacts existing in the restored image locally, it is necessary to refine the obtained coarse transmittance map. Because the traditional transmittance map optimization algorithm has poor edge retention ability and serious loss of detail, this paper proposes an adaptive weighted guidance filtering algorithm based on the traditional algorithm. By adding an adaptive weighted factor, the purpose of further improving the edge detail of the fine transmittance map is further achieved.Result In this paper, the general dehazing test image and the foggy image taken by the small UAV are taken as experimental objects. By comparing and analyzing the restoration effects of the four combined steps algorithms, the rationality of the improved methods and the superiority of the overall algorithm are verified. The experimental results show that the mixed prior theory improves the distortion of dark priori in bright region and the deficiency of color attenuation priori in dense fog processing, and achieves better visual effect. Weighted guided filtering improves the image edge blurring and makes the image edge details clearer after restoration. Compared with other comparison algorithms, the proposed algorithm has better visual effects and the edge details of the haze-free images are more obvious. The average of comprehensive evaluation index increases is larger.Conclusion For the restoration of hazy images, the superiority of the improved model proposed in this work is demonstrated through theoretical analysis and experimental verification. Restoration through the proposed algorithm is better than that with the traditional algorithm. The main conclusions are as follows: Mixed prior theory can improve the problem of prior blind area in the dark primary color prior theory and color attenuation theory to a certain extent, and the effect of image de-fogging is better. Adaptive guided filtering algorithm can optimize the transmittance image better and improve the edge sharpness of the defogging image. Combining the hybrid prior theory with the adaptive guided filtering algorithm, the image de-fogging effect can be improved. under the same conditions, compared with the traditional fog removal algorithm, the algorithm proposed in this paper has better image restoration effect. In this paper, there are still some shortcomings in the parameter setting of regional segmentation, only adjusting parameters through experience, lack of certain theoretical basis, the next step will be to further study the parameter setting. MPWGF has broad application prospects in image restoration, artificial intelligence, photogrammetry and other fields.
Keywords
QQ在线


订阅号|日报