Current Issue Cover


摘 要
目的 针对水下偏振图像存在雾状模糊和场景细节不明显的问题,以水体透射率图与目标反射光图像存在的相互独立性为基础,提出了一种基于结构相似性的水下偏振图像复原方法,旨在提高水下偏振图像的清晰度、对比度和色彩真实度。方法 首先,获取同一水下场景下具有正交偏振方向且分别具有最大和最小光强的两幅偏振图像;然后根据透射率图与目标反射光之间的统计无关性,使用结构相似性推导求解透射率的关系式,并通过偏振差分图像计算透射率的初始值,利用该关系式进行水体透射率的迭代求解;最后将透射率代入偏振成像模型得到目标反射光图像,进而进行颜色校正得到复原图像。结果 选取多组正交的水下偏振图像作为研究对象,采用本文提出的方法与另两种偏振复原算法对其进行复原处理,使用对比度、信息熵、灰度平均梯度、峰值信噪比、增强量以及时间等量化指标进行评估。对比实验结果表明,本文算法在对比度、信息熵、灰度平均梯度、增强量以及颜色恢复上都优于另两种偏振图像复原方法,复原效果有较大幅度的提有待提高,但算法运行时间较长,实时性有待提高。结论 本文基于水下偏振成像模型的分析以及透射率图与目标反射光图像之间的统计无关性,从水体透射率的估计出发进行图像复原,有效地解决了水下偏振图像细节模糊、对比度低的问题。通过对算法实验效果的主客观分析表明,本文算法能有效地复原水下偏振图像,得到对比度高、细节明显和色彩丰富的恢复图像。
Underwater polarized images restoration algorithm based on structural similarity

fan xinnan,chen jianyue,zhang xuewu,shi pengfei,zhang zhuo(College of Internet of Things Engineering, Hohai University)

Objective The restoration algorithms for single image are numerous, and they are remarkable for the defogging of sky images, but most of them can not be directly applied to the restoration of underwater images. The goal of image restoration is to process the degenerated images to recover to the ideal image before degeneration. Because of underwater insufficient illumination and unevenly distributed light, the results obtained by restoration method of single image are affected by light variation. Generally, the results are not satisfactory. The polarization is the basic feature of light and the reflected light of the underwater objects are mostly partial polarized, so the underwater polarized images have special polarization characteristic, and the underwater images restoration based on multiple polarized images has gradually become hot research area in recent years. Aimed at the mistiness and unobvious details of underwater polarized images, a restoration method of underwater polarized image based on structural similarity is proposed, which is expected to improve the clarity, contrast and color fidelity of images. Method Firstly, A couple of images taken through a polarizer at orthogonal orientations are obtained. The images have the best and the worst backscatter respectively. Then, since the water transmittance is only related to the depth of field and the attenuation coefficient of the water body, but the object radiance depends on the incident light and the surface characteristics of the object, so we can assume that they are mutually independent. The structural similarity can measure the similarity of two images from brightness, contrast and structure, and it can directly describe the correlation between the two images. Secondly, based on the irrelevance relationship between the transmittance and the object radiance, the solution formula of water transmittance is derived by the structural similarity. The difference of the two polarized images is also the difference of the background light in images, and it is also the function of depth of field, so the polarized-difference image is used for calculating the initial value of the transmittance during the iteration solving process of it. An accurate transmittance is necessary for a good restoration of images. Finally, the object radiance is obtained by inversing the underwater polarization imaging model, and the color correction is carried out for it to obtain the restored image. The color correction based on single point chooses the point whose color information is kept well as the reference pixel, and then normalize the global pixels by the reference pixel to realize the color correction of the whole image. Result In the experiment, the proposed algorithm is compared with two other polarized restoration algorithms to test its effectiveness, and several groups of underwater polarized are selected as research objects. The images used in the paper are obtained from the relevant literatures. Quantitative indicators such as contrast, information entropy, GMG(gray mean grads), PSNR(peak signal-to-noise ratio), EME(the measure of enhancement) and runtime are used for evaluation of effect. The result shows that the contrast, information entropy and GMG of our method are better than other two algorithms. And a greater improvement of restoration effect is achieved. The YY algorithm removes the blur of original images to some extent, but supersaturation phenomenon exists in some object areas of recovered images. Due to the inaccurate estimation of the degree of polarization of the object radiance, the image restored by the Huang algorithm are generally too dark to identify the details of scene. The comparison of evaluation parameters shows that the contrast and GMG of our method are twice as high as YY algorithm. And the color distribution of images recovered by our method are more homogeneous, which makes the images have plentiful enough information and the highest information entropy. The prominent EME also proves that our result have clear texture, high contrast and well restoration. Since some color channels of the images obtained by the Huang algorithm are not recovered, its color tones are single and the value of some color channels are as low as the raw images, result in the small MSE(mean square error) and too high PSNR. On the time cost, our method and Huang algorithm are relatively longer than YY on account of the traversal process of parameters. Conclusion Based on the analysis of underwater polarization imaging model and the statistical independence relationship exists in the object radiance and water transmittance, the image restoration is carried out after the estimation of transmittance successfully. And the problem of blurred details and low contrast in underwater polarized image are effectively solved. The result of subjective and objective analysis show that the proposed algorithm can recover the underwater polarized images effectively, and obtain the restored images with high contrast, obvious details and rich color. Compared to other algorithms, the proposed algorithm can improve the contrast, clarity and color balance of underwater polarized images more greatly, which provides an important foundation for underwater target recognition and analysis.