一种适用于人脸检测的自适应光照补偿方法

为了解决复杂光照条件下的人脸检测问题,提出一种人脸光照补偿新方法。该方法先使用高通滤波增强边缘信息,同时利用对数变换和指数变换调节全局亮度,最后利用非线性变化削弱局部高光和阴影的影响,改善图像光照不均衡的情况,最终实现光照补偿。在YaleB人脸库、Orl人脸库以及自建人脸库上分别对光照不均匀人脸图像和均匀光照下的人脸图像进行了实验,证明该方法能有效地进行光照补偿,提高人脸检测率。     

结合边缘信息和门卷积的人脸修复算法

针对任意形状遮挡下人脸修复,现有方法容易产生边缘模糊和恢复结果失真等问题。提出了一种结合边缘信息和门卷积的人脸修复算法。首先,通过先验人脸知识产生遮挡区域的边缘图,以约束人脸修复过程。其次,利用门卷积在部分像素缺失下的精确局部特征描述能力,设计面向图像修复的门卷积深度生成对抗网络(GAN)。该模型由边缘连接生成对抗网络和图像修复生成对抗网络两部分组成。边缘连接网络利用二值遮挡图和待修复图像及其边缘图的多源信息进行训练,实现对缺失边缘图像的自动补全和连接。图像修复网络以补全的边缘图为引导信息,联合遮挡图像进行缺失区域修复。实验结果表明:相比其他算法,该算法修复效果更好,其评价指标比当前基于深度学习的图像修复算法更优。     

基于光照方向估计的人脸图像阴影补偿

光照变化是影响人脸识别率的关键问题之一.人脸图像中的阴影严重影响了光照不变特征的有效提取.采用了一种基于光照方向估计的阴影补偿方法.因为人脸形状的相似性,某一点光源方向产生的阴影特性具有相似性.利用这一特性,首先通过子空间方法确定人脸图像的光源方向,然后再对该人脸图像进行相应的阴影补偿.在对人脸库YaleB和Extended YaleB （38人脸）的实验中表明,64种光源方向的识别率为96.8％.该方法能有效消除阴影便于后续光照不变特征的有效提取.     

一种基于遮挡区域补偿的眼镜去除方法

为了提高戴眼镜人脸图像的识别率, 提出了一种从人脸图像中检测并去除眼镜的方法。首先对输入的戴眼镜人脸图像与系统预留的无眼镜人脸图像进行基于人眼位置的标定, 检测出眼镜遮挡区域, 再用无眼镜人脸图像中对应的遮挡区域对戴眼镜人脸图像进行补偿, 从而合成了对应输入图像的不戴眼镜的人脸图像。实验结果表明, 该方法能有效地合成无眼镜人脸图像, 将合成后的人脸图像再应用于人脸识别系统, 识别率显著提高。     

遮挡条件下实时人脸检测

针对遮挡条件下人脸检测召回率不高和训练样本不足的问题,提出了一种数据增强的方法。计算训练样本的平均脸图像,对样本进行手动遮挡处理;使用平均脸策略和图像分割理论对遮挡区域进行分割和处理,获得类似于自然环境下的遮挡样本;使用处理后的样本训练Ada Boost人脸检测器。实验结果表明:通过该方法训练的人脸检测器能够有效地提高遮挡条件下的人脸召回率,同时可以达到实时检测的效果。     

人脸彩色图像光照校正方法研究

对人脸彩色图像的高光和阴影部分进行研究.对于高光区域,利用人脸肤色与像素空间分布特征进行检测,根据在rgb空间中各分量颜色偏移率实现高光区域自动校正,通过边缘拟合方法校正超出显示范围的图像.对于阴影区域,采用Retinex方法进行光照增强.实验结果表明,该方法处理速度快,且能提高图像对比度.     

人脸遮挡识别技术在智能备件柜身份认证中的应用

研究人脸遮挡识别技术在智能备件柜身份安全认证中的应用，以提高智能备件柜的安全性。依据人脸肤色以及形状特征定位人脸位置并对图像进行灰度归一化处理，采用稀疏编码方法将遮挡人脸图像的局部特征转变为特征向量，运用支持向量机构建遮挡人脸图像分类器，根据输出结果判断实时采集的人脸与数据库人脸图像是否匹配，实现智能备件柜的身份安全认证。实验结果表明，该方法能够准确识别遮挡人脸完成身份安全认证，人脸识别的误识率低，识别速度快，可保障智能备件柜安全。     

一种能实现免脱帽人脸识别系统算法

针对人脸识别系统在人脸被遮挡情况下识别率低的问题,为进一步提升人脸在遮挡情况下的识别率,文章提出一种通过图像多方向梯度值,使用融合、补偿等方式产生可以对原图像进行特征描述的特征图像,通过对特征图进行一系列处理后实现人脸识别的算法;算法首先计算图像四方位的梯度值;其次对4个梯度值进行融合运算,产生合融梯度、差融梯度;再次以合融梯度、差融梯度作为补偿变量在原图像上进行适当系数的补偿,形成人脸图像特征图;然后对特征图依次进行直方图统计、主成分分析后,使用SVM分类器进行分类识别;使用Matlab2016试验仿真平台在ORL、CMU_PIE等多个人脸数据库上进行测试,分别取得100%、92.21%的准确率,结果表明推荐算法在人脸被遮挡情况下的识别率具有很好的表现。     

复杂光照环境下视频人脸序列的自动检测方法

基于视频序列人脸自动检测是人脸跟踪、识别等研究的基础.提出了一种结合图像增强技术、gabor特征变换和adaboost算法的视频序列人脸检测方法,其主要思想是使用图像增强技术对图像进行光照补偿,减轻不同的光照条件(如局部的阴影和高亮等)对检测结果的影响.该方法首先通过高频增强滤波强化图像的边缘和细节信息,用基于直方图的技术采调节图像的亮度,然后应用gabor小波变换进行特征抽取,最后采用adaboost方法训练样本,完成人脸的检测.实验表明,该方法能够在不同的光照条件下准确检测出人脸,显示出较强的鲁棒性.     

基于鲁棒主成分分析的人脸子空间重构方法

子空间方法是人脸识别中的经典方法,其基本假设是人脸图像处于高维图像空间的低维子空间中.但是,由于光照变化、阴影、遮挡、局部镜面反射、图像噪声等因素的影响,使得子空间假设难以满足.为此,提出一种基于鲁棒主成分分析的人脸子空间重构方法.该方法将人脸图像数据矩阵表示为满足子空间假设的低秩矩阵和表征光照变化、阴影、遮挡、局部镜面反射、图像噪声等因素的误差矩阵之和,利用鲁棒主成分分析法求解低秩矩阵和误差矩阵.实验结果表明,文中方法能够有效地重构人脸图像的低维子空间.     

结合区域配对的室外阴影检测

目的 针对现有大多数阴影检测算法在检测细长阴影、自阴影、区分阴影与暗色像素等方面的不足,提出一种新的结合区域配对的阴影检测算法.方法 首先通过均值漂移算法和canny检测算法,分割图像得到每个独立的区域;然后从每个区域中提取纹理和亮度建立单个区域的阴影模型,再从区域对中提取纹理直方图的距离、颜色比(分别在RGB和Lab空间下)以及HSI空间下H和I两通道的比值等特征建立区域对的阴影模型;最后根据上述两个模型运用图割理论检测阴影.结果 实验结果表明,本文算法在阴影检测上的准确率高达85.2%,远高于其他算法,检测速度也比其他算法快34%左右.该算法不仅能有效地检测细长阴影和自阴影,还能较好地区分阴影与暗色像素.结论 提出了一种新的阴影检测算法,通过区域配对的方法实时处理单幅室外图像.实验结果表明,该算法在检测细长阴影、自阴影以及区分阴影与暗色像素等方面有良好的效果.     

Image shadow removal using pulse coupled neural network

This paper introduces an approach for image shadow removal by using pulse coupled neural network (PCNN), based on the phenomena of synchronous pulse bursts in the animal visual cortexes. Two shadow-removing criteria are proposed. These two criteria decide how to choose the optimal parameter (the linking strength /spl beta/). The computer simulation results of shadow removal based on PCNN show that if these two criteria are satisfied, shadows are removed completely and the shadow-removed images are almost as the same as the original nonshadowed images. The shadow removal results are independent of changes of intensities of shadows in some range and variations of the places of shadows. When the first criterion is satisfied, even if the second criterion is not satisfied, as to natural grey images that have abundant grey levels, shadows also can be removed and PCNN shadow-removed images retain the shapes of the objects in original images. These two criteria also can be used for color images by dividing a color image into three channels (R, G, B). For shadows varying drastically, such as the noisy points in images, these two criteria are still right, but difficult to satisfy. Therefore, this approach can efficiently remove shadows that do not include the random noise.     

基于区域与光照不变性的运动阴影检测算法

在户外的视频监控系统中，运动目标的阴影降低了系统对目标识别与跟踪的能力。传统的基于像素的阴影检测算法易受噪声的影响。为了提高阴影检测算法的准确性,提出了一种基于区域与光照不变性的运动阴影检测算法。该算法从阴影的物理特性出发,考虑了区域内像素的总体特征。将运动区域采用EM聚类算法进行分块,对其中的小块向邻近的大块进行合并。对其中的每一块，根据阴影区域和相对应的背景区域之间的光照不变性进行阴影检测。实验结果表明，该算法能够很好地抑制噪声，准确地检测出阴影，明显比基于像素的算法有效。     

结合多种特征的高分辨率遥感影像阴影检测

针对现有的阴影检测算法对较亮阴影和较暗地物中的阴影不能同时较好地检测等问题, 提出一种结合多种特征的高分辨率遥感影像阴影检测方法.该算法首先结合主成分分析、颜色特征和直方图的分割构建多种阈值检测条件, 然后综合多种特征来进行遥感影像阴影的初步检测, 最后通过分析RGB模型在阴影与非阴影地物上的差别, 利用颜色特性最终检测出阴影区域.实验结果表明, 本文算法能有效检测较亮阴影和较暗地物中的阴影.与现有方法相比, 较亮阴影的平均总错误率从水平集法的31.85%降至24.61%, 较暗地物中阴影的平均总错误率从自动检测法的37.75%降至23.30%.     

Shadow Segmentation and Classification in a Constrained Environment

A shadow identification and classification method for real images is developed in this paper. The method is based on the extensive analysis of shadow intensity and shadow geometry in an environment with simple objects and a single area light source. The procedure for identifying shadows is divided into three processes: low level, middle level, and high level. The low level process extracts dark regions from images. Dark regions contain both shadows and surfaces with low reflectance. The middle level process performs feature analysis on dark regions, including detecting vertices on the outlines of dark regions, identifying penumbrae in dark regions. classifying the subregions in dark regions as self-shadows or cast shadows, and finding object regions adjacent to dark regions. The high level process integrates the infonnation derived from the previous processes and confirms shadows among the dark regions.     

Landsat 8 Data Mountain Shadow Removal based onBrightness Recovering Model

Mountain region in remotely sensed imagery are usually covered by shadows，which reduce the accuracy of information extraction.Therefore，in this paper a method based on intensity restoration is putting forward necessarily.First，Shadow Detection (SD) was constructed by the Max function and the band ratio to identify shadows.Thus，mountain shadows were extracted combined with the slope factor and SD，through the grid randomly arranged verification point verification accuracy.Second，the intensity curve model of the shadow area was fitted by ground data of the shadow and the transition rules of pixel intensity from the shadow to non\|shaded area.Third，the intensity restoration model was established by the derivative function of intensity curve to remove shadows.The results of the model on Changting Landsat 8 imagery indicated the extraction accuracy of the mountain shadow was 99.06% and the Kappa coefficient was 98%；According to the cluster analysis，the restoration and non\|shaded samples were the same type；Processed by the intensity restoration model，the average intensity of the shadow was increased by 13%，and the standard deviation was reduced by 80% and the clustering distances was reduced by 96%.respectively，average intensity of the shadow increased by 6.7%，the standard deviation was reduced by 73.7% and the clustering distances was reduced by 88.3% when compared with ATCOR_3，and average intensity of the shadow reduced by 1.8%，the standard deviation was increased by 6.7% and the clustering distances was reduced by 90% when compared with unitary linear restoration model.In the process of removing the mountain shadows，the intensity restoration method is neither replacing the shaded pixels nor interference with non\|shaded pixels and could preserve the spectral and intensity characteristics of shaded pixels better.  ；     

De‐shadowing of satellite/airborne imagery

A de‐shadowing technique is presented for multispectral and hyperspectral imagery over land acquired by satellite/airborne sensors. The method requires a channel in the visible and at least one spectral band in the near‐infrared (0.8–1?µm) region, but performs much better if bands in the short‐wave infrared region (around 1.6 and 2.2?µm) are available as well. The algorithm consists of these major components: (i) calculation of the covariance matrix and zero‐reflectance matched filter vector, (ii) derivation of the unscaled and scaled shadow function, (iii) histogram thresholding of the unscaled shadow function to define the core shadow areas, (iv) region growing to include the surroundings of the core shadow areas for a smooth shadow/clear transition, and (v) de‐shadowing of the pixels in the final shadow mask. The critical parameters of the method are discussed. Example images from different climates and landscapes are presented to demonstrate the successful performance of the shadow removal process over land surfaces.     

Shadow detection in colour high‐resolution satellite images

Image shadow segmentation has become a major issue in satellite remote sensing because of the recent commercial availability of high‐resolution images. Detecting shadows is important for successfully carrying out applications such as change detection, land monitoring, object recognition, scene reconstruction, colour correction, etc. This paper presents a simple and effective procedure to segment shadow regions on high‐resolution colour satellite images. The method applies a region growing process on a specific band (namely, the c ₃ component of the c ₁ c ₂ c ₃ colour space). To gain in robustness and precision, the region expansion also imposes a restriction on the saturation and intensity values of the shadow pixels, as well as on their edge gradients. The proposed method has been successfully tested on QuickBird images acquired under different lighting conditions and covering both urban and rural areas.     

融合色彩比和梯度不变性的运动阴影检测

为解决运动前景的准确分割受运动阴影影响的问题,提出了一种融合色彩比和梯度不变性的运动阴影检测算法。该算法分析了阴影像素的色彩比和区域纹理梯度的光照不变性,利用亮度变化特性和色彩比不变性初步确定候选运动前景中的阴影像素,然后在候选阴影区域利用纹理梯度不变性进行去错处理,两者的结合弥补了单一特征或单一类型特征的阴影检测性能差的缺陷,提高了阴影检测率和阴影分辨率,能够准确地将阴影和前景区别开来。     

Efficient detection under varying illumination conditions and image plane rotations

This paper focuses on the detection of objects with a Lambertian surface under varying illumination and pose. We offer to apply a novel detection method that proceeds by modeling the different illuminations from a small number of images in a training set; this automatically voids the illumination effects, allowing fast illumination invariant detection, without having to create a large training set. It is demonstrated that the method “fits in” nicely with previous work about modeling the set of object appearances under varying illumination. In the experiments, an object was correctly detected under image plane rotations in a 45° range, and a wide variety of different illuminations, even when significant shadows were present.