基于互结构正则约束的红外偏振图像增强算法

为有效改善红外偏振图像视觉效应,提高红外偏振成像质量,提出了基于互结构正则约束的红外偏振图像增强算法。根据红外偏振特性描述,对Stokes参数Q分量与U分量进行加权邻域梯度融合,获得起偏特征图像,捕获目标边缘、轮廓的偏振特性;提出互结构正则约束模型,以梯度幅值相似算子联合正则约束融合结果与起偏特征图像的边缘结构相似性,及与辐射强度图像的灰度一致性,优化得到增强后的高质量红外偏振图像。实验结果表明,基于互结构正则约束的红外偏振图像增强算法,能有效提高红外偏振图像对比度与清晰度,同时提升复杂背景下人造目标边缘轮廓的偏振显著性,算法快速,工程实时性高.     

基于偏振测量的雾天图像场景分割

现有场景分割方法主要依赖于图像亮度、颜色和纹理等特征,然而在雾天图像中提取这些特征将变得困难且不稳定.基于此本文提出了适用于雾天图像场景分割的特征矢量,以及相应的特征提取算法.特征矢量由目标偏振度、深度和颜色三部分组成.特征提取算法分别为:用去相关的方法从图像偏振度分离出大气偏振度和目标偏振度;根据雾天退化模型和雾天图像偏振表示形式推导出场景深度信息;利用两幅偏振图像求出非偏振彩色图像,从而得到场景的颜色信息.将这些特征构成的特征矢量用于基于图的分割算法中,并从两个方面比较了仅使用颜色特征和使用本文特征矢量的分割结果.最后得出结论:对于雾天图像而言,这些特征比通常的颜色特征更加有效和鲁棒.     

基于光照-反射成像模型和形态学操作的多谱段图像增强算法

为解决多谱段降质图像增强问题,提出了一种基于光照-反射成像模型和形态学操作的多谱段图像增强算法.首先对图像饱和度使用自适应非线性拉伸函数进行拉伸,使增强后的图像色彩更加饱和、自然;接下来利用引导滤波算法提取出图像的光照分量,提出了一种基于细节特征的加权融合策略,利用光照分布特性构造了一种自适应Gamma校正函数对光照分量进行处理,并将其与利用对比度受限的自适应直方图均衡化方法处理后的光照分量以及原始光照分量进行融合;然后在反射分量校正时,构造了一种形态学操作函数来校正反射信息;最后合并光照分量和反射分量,并与处理后的饱和度分量和色调分量一起得到增强图像.采用主客观评价指标对可见光低照度图像、水下图像、高动态范围图像、沙尘暴图像、雾天图像和热红外图像6种降质多谱段图像实验结果进行分析比较,结果表明本文算法能够有效地抑制图像噪声、增强图像细节信息、改善图像视觉效果,可应用于多种图像增强领域.     

多算法融合的自适应图像增强方法

提出一种多算法融合的图像增强方法,用于工程应用中的复杂降质图像的细节特征恢复.该方法汲取了Laplacian变换法、Sobel梯度法、盒状滤波法、非锐化掩蔽法及灰度幂律法等算法的优点,可对模糊图像进行自适应增强.通过拉普拉斯滤波器和梯度滤波器将原始图像分为基础层、细节层及边缘特征层;对微小细节信息及边缘特征信息进行增强,对基础信息进行压缩;然后采用盒装滤波器对图像的三个分层进行平滑过度及噪音过滤,最后使用非锐化掩蔽法和灰度变换来增加图像灰度的动态范围,从而得到增强后的图像.在相同的工况下,该方法分别与直方图均衡法、自适应伽马矫正法及小波变换的图像增强法实验结果进行对比,结果表明,该方法将图像的清晰度提高了13.1%~126.1%,能有效地处理复杂型感染的图像,避免图像过度增强,可以获得适合人眼的最佳视觉细节内容的增强效果.     

A Crop Image Segmentation and Extraction Algorithm Based on Mask RCNN

The wide variety of crops in the image of agricultural products and the confusion with the surrounding environment information makes it difficult for traditional methods to extract crops accurately and efficiently. In this paper, an automatic extraction algorithm is proposed for crop images based on Mask RCNN. First, the Fruits 360 Dataset label is set with Labelme. Then, the Fruits 360 Dataset is preprocessed. Next, the data are divided into a training set and a test set. Additionally, an improved Mask RCNN network model structure is established using the PyTorch 1.8.1 deep learning framework, and path aggregation and features are added to the network design enhanced functions, optimized region extraction network, and feature pyramid network. The spatial information of the feature map is saved by the bilinear interpolation method in ROIAlign. Finally, the edge accuracy of the segmentation mask is further improved by adding a micro-fully connected layer to the mask branch of the ROI output, employing the Sobel operator to predict the target edge, and adding the edge loss to the loss function. Compared with FCN and Mask RCNN and other image extraction algorithms, the experimental results demonstrate that the improved Mask RCNN algorithm proposed in this paper is better in the precision, Recall, Average precision, Mean Average Precision, and F1 scores of crop image extraction results.     

Fusion of visible and infrared images using saliency analysis and detail preserving based image decomposition

Image fusion for visible and infrared images is a significant task in image analysis. The target regions in infrared image and abundant detail information in visible image should be both extracted into the fused result. Thus, one should preserve or even enhance the details from original images in fusion process. In this paper, an algorithm using pixel value based saliency detection and detail preserving based image decomposition is proposed. Firstly, the multi-scale decomposition is constructed using weighted least squares filter for original infrared and visible images. Secondly, the pixel value based saliency map is designed and utilized for image fusion in different decomposition level. Finally, the fusion result is reconstructed by synthesizing different scales with synthetic weights. Since the information of original signals can be well preserved and enhanced with saliency extraction and multi scale decomposition process, the fusion algorithm performs robustly and excellently. The proposed approach is compared with other state-of the-art methods on several image sets to verify the effectiveness and robustness.     

基于色调恒定的MSR图像增强算法

由于多尺度Retinex算法增强后图像存在细节信息减弱和颜色失真等不足,本文提出了一种色调恒定的图像增强算法.在原图像中去掉用多尺度高斯函数估计的光照分量,结合参量自适应的非线性函数调整亮度,依据色调恒定的理论保持增强后图像的颜色.与多尺度Retinex比较的实验结果表明,本文算法更有效,增强后的图像不仅细节清晰,而且色彩自然、不失真且运行速度快.     

基于Retinex理论与概率非局部均值的红外图像增强方法

针对传统红外图像增强算法中细节模糊及过度增强的问题,提出了一种基于Retinex理论与概率非局部均值相结合的红外图像增强方法.首先通过单尺度Retinex方法调整图像中过暗与过亮部分的灰度级;然后利用概率非局部均值对图像进行分解处理得到基本层与细节层,对基本层采用直方图均衡化拉伸对比度,对细节层采用非线性函数进行增强;最后,将不同层次的结果融合得到对比度与细节增强的红外图像.用该方法对多组不同场景的红外图像进行仿真实验,并将其与多种增强方法进行主、客观对比分析,结果表明所提方法在红外图像的细节及对比度增强方面都获得了更好的效果.     

基于潜在低秩表示与复合滤波的红外与弱可见光增强图像融合方法

针对传统红外与弱可见光图像融合算法中存在的亮度与对比度低、细节轮廓信息缺失、可视性差等问题,提出一种基于潜在低秩表示与复合滤波的红外与弱可见光增强图像融合方法.该方法首先利用改进的高动态范围压缩增强方法增强可见光图像提高亮度;然后利用基于潜在低秩表示与复合滤波的分解方法分别对红外与增强后的弱可见光图像进行分解,得到相应的低频和高频层;再分别使用改进的对比度增强视觉显著图融合方法与改进的加权最小二乘优化融合方法对得到的低频和高频层进行融合;最后将得到的低频和高频融合层进行线性叠加得到最终的融合图像.与其他方法的对比实验结果表明,用该方法得到的融合图像细节信息丰富,清晰度高,具有良好的可视性.     

A joint image encryption and watermarking algorithm based on compressive sensing and chaotic map

In this paper,a compressive sensing(CS) and chaotic map-based joint image encryption and watermarking algorithm is proposed.The transform domain coefficients of the original image are scrambled by Arnold map firstly.Then the watermark is adhered to the scrambled data.By compressive sensing,a set of watermarked measurements is obtained as the watermarked cipher image.In this algorithm,watermark embedding and data compression can be performed without knowing the original image;similarly,watermark extraction will not interfere with decryption.Due to the characteristics of CS,this algorithm features compressible cipher image size,flexible watermark capacity,and lossless watermark extraction from the compressed cipher image as well as robustness against packet loss.Simulation results and analyses show that the algorithm achieves good performance in the sense of security,watermark capacity,extraction accuracy,reconstruction,robustness,etc.     

基于小波域三状态HMT模型的含噪图像增强

针对含噪图像增强问题,提出一种基于小波域三状态隐马尔可夫树模型的方法,采用三状态的高斯混合模型逼近小波系数的分布,不需要设定精确的阈值,依据期望最大算法训练得到的每个系数所属状态的后验概率,将系数区分为噪声系数、弱边缘系数和强边缘系数,然后通过抑制噪声系数,增强细节特征系数来达到对含噪图像增强的目的,并引入循环平移策略避免人工失真.通过对含噪的标准图像和人脑核磁共振图像进行仿真实验,并与几种经典的图像增强方法作视觉上的对比和定量分析.实验结果表明,本文所提出的方法具有很好的鲁棒性,在突出了图像中更多的细节信息的同时,可以有效抑制噪声.     

基于光饱和影响校正的作物叶绿素分布光谱成像检测

叶绿素含量是作物光合能力与营养评价的重要指标,因此快速检测作物叶绿素含量与分布可为作物营养动态分析与长势评估提供支持。基于RGB(Red, Green, Blue)和NIR(Near Infrared)多光谱图像的获取,开展玉米作物营养状态分布光谱学成像检测。构建了多光谱图像采集平台获取RGB和NIR图像,研究了基于光饱和校正算法的RGB图像的光饱和校正与NIR图像去噪方法,通过图像的匹配分割,冠层的提取校正,建立了基于冠层图像的作物SPAD值检测模型与分布成图。采集15株玉米植株RGB-NIR图像,并同步获取不同植株,不同位置共68个叶绿素含量指标SPAD值。首先对RGB图像进行光饱和校正,再对NIR图像进行滤波和图像增强,其次对RGB和NIR图像进行了SURF(speeded-up robust features)和RANSAC(random sample consensus)图像匹配,利用RGB图像的颜色特征,采用ExG(Extra Green)和OTSU算法生成分割掩模,对RGB图像和NIR图像进行分割提取,提取图像的R, G, B和NIR分量,利用4阶灰度板进行反射率校正,然后计算作物图像中像素级P_SPAD值,并建立图像P_SPAD值与叶绿素仪SPAD值的拟合模型,最后绘制作物SPAD分布图。通过HSI(Hue, Saturation, Intensity)彩色模型中的I分量直方图对比去饱和前后光分布范围,以作物SPAD值分布图验证光饱和校正算法对作物叶绿素含量分布检测提升的效果。RGB图像光饱和校正前I分量集中在[140～180]之间,光饱和校正后的RGB图像I分量集中在[85～130]之间, 校正了相机成像时产生模糊和RGB图像饱和。对分割后的RGB图像和NIR图像提取R,G,B,NIR分量进行4阶灰度板校正,相关系数分别为0.829,0.828,0.745和0.994,进而生成R,G,B和NIR四波段的反射率伪彩色图像,反射率R_NIR＞R_C＞R_R＞R_B。体现了作物的在蓝光和红光区域吸收光,在绿光区域和近红外区域反射光的光谱特性。校正前后的R和NIR分量反射率计算图像P_SPAD值拟合叶绿素含量指标SPAD值的模型结果显示,校正前R2为0.332 6,校正后R2为0.619 3,绘制作物的SPAD特征分布图,可为作物的营养动态快速分析与分布检测提供技术支持。     

基于Contourlet变换和形态学的图像增强方法

传统的图像增强算法在增强图像时,增大噪音,同时丢失细节.针对该问题,结合Contourlet变换中相关系数理论,提出了基于Contourlet变换和数学形态学的图像增强新算法.首先,对图像进行Contourlet变换分解,采用数学形态学算子对高频细节部分区分为细节信息和噪音产生的系数,然后,对变换系数采用非线性映射函数进行增强.最后,利用修改后的变换系数进行Contourlet逆变换得到增强后的图像.实验表明,该方法无论是增强效果还是抗噪性能都明显优于传统的图像增强算法.     

Fusion for visible and infrared images using visual weight analysis and bilateral filter-based multi scale decomposition

Fusion for visible and infrared images has been an important and challenging work in image analysis. Both the feature information in infrared image and abundant detail information in visible image should be preserved and enhanced in fused result. In this paper, a detail enhanced fusion algorithm through visual weight analysis based on smooth-inspired multi scale decomposition is proposed. With variable parameter, bilateral filter-based idea successfully decomposes the two source image into several scales. At each scale level, visual weight map is calculated and used for fusion. Finally, those levels are synthetized with proper weights. Using this idea, the detail information could be enhanced easily. The experimental results demonstrate the proposed approach performs better than other methods, especially in visual effect and keeping details.     

基于可见光的多波段偏振图像融合新算法

采用了一种新的基于小波变换的偏振图像融合算法.首先,将两个波段中的每一波段三幅偏振图像利用小波变换分解成低频和高频部分,低频的小波系数平均值作为融合后的低频系数,高频细节系数根据不同区域特征选择方法以及对应输入图像小波系数的窗口区域方差来确定融合后高频小波系数,得到一个波段一幅图像.接着,将得到的图像再进行小波分解,采用低频图像的小波系数最小值作为融合后的低频系数,高频图像根据纹理一致性测度的纹理检测确定融合规则,用来调整高频小波系数,将来自不同图像的特征与细节融合在一起,并对融合图像质量进行了对比评价.实验结果表明,融合后的偏振图像不仅反映了场景的偏振信息,而且还包含了丰富的光谱信息,目标与背景的衬比度也得到了增强,为进一步的目标检测和识别提供了便利.     

一种基于图像局部信息的红外图像动态范围压缩算法

红外焦平面探测器输出的信号通常为14bit数字图像,要在常用的显示器上显示必须将14bit的图像压缩到8bit,压缩算法直接影响到了图像的细节和对比度。提出了一种基于图像局部信息的红外图像动态范围压缩算法,通过高斯滤波器将原始红外图像分解为基图像和细节图像,并统计原始图像每个像素的局部方差,然后根据每个像素的局部方差计算细节图像的自动增益函数,最后,将基图像进行平台直方图均衡化后与细节图像合成得到压缩后的图像。通过对多种场景的红外图像进行仿真验证,表明该算法能较好地增强图像的细节。     

基于BIRCH聚类加速的彩色图像增强算法

针对现存的大多图像增强算法增强的图像可见性丢失问题,提出了一种基于BIRCH聚类加速的彩色图像增强算法;首先,通过BIRCH聚类加速确定数据库中与输入图像直方图相似度最高的图像来提取图像特征;然后,选择最小欧氏距离的特征值进行图像融合以获取目标图像;最后,增强图像通过目标图像直方图规范化和后期处理获得;大量图像融合实验结果验证了算法的有效性,该算法扩展了图像增强的类别,解决了增强过程中可能出现的可见性丢失问题,使图像增强的适应性更强;另外,EM、CII和SSIM评估指标的结果表明该算法明显改善了增强效果。     

A polynomial hybrid reflection model and measurement of its parameters based on images of sample

Reflectance model is a basic concept in computer vision. Some existing models combining the classical diffuse reflectance model and those for surfaces containing specular components can approximately describe real reflectance. But the ratio of diffuse and specular reflection decided manually has no clear meaning.We propose a new polynomial hybrid reflectance model. The reflectance map equation with a known shape (for example cylinder) as a sample is used to estimate parameters of the proposed reflectance model by least square regression algorithm. Then the reflectance parameters for surfaces of the same class of materials can be determined. Experiments are performed for a metal surface. The synthesis images produced by the proposed method and existing ones are compared with the real acquired image, and the results show that the proposed reflectance model is suitable for describing real reflectance.     

高光谱技术融合图像信息的牛肉品种识别方法研究

高光谱图像包含了大量的光谱信息和图像信息,采用高光谱成像技术对牛肉品种进行识别。获取可见-近红外(400~1000 nm)光谱范围内的安格斯牛、利木赞牛、秦川牛、西门塔尔牛、荷斯坦奶牛五个品种共252个牛肉样本的高光谱图像。在ENVI软件中对高光谱图像进行阈值分割并构建掩膜图像,获取样本的感兴趣区域(ROI),并结合伪彩色图对牛肉样本的反射率指数进行可视化表达;采用Kennard-Stone(KS)法对样本集进行划分以提高模型的预测性能;对原始光谱采用卷积平滑(SG)、区域归一化(Area normalize)、基线校正(Baseline)、一阶导数(FD)、标准正态变量变换(SNV)及多元散射校正(MSC)等6种方法进行预处理;采用竞争性自适应重加权算法(CARS)提取特征波长。然后利用颜色矩对不同牛肉样本的颜色特征进行提取;对原始光谱图像进行主成分分析,结合灰度共生矩阵(GLCM)算法,提取主要纹理特征。最后结合偏最小二乘判别(PLS-DA)算法建立牛肉样本基于特征波长、颜色特征以及纹理特征的识别模型。KS法将牛肉样本划分为校正集190个,预测集62个;将未经预处理的光谱数据与经过6种不用预处理的光谱数据进行建模分析,结果发现经FD法处理后的光谱数据所建模型的识别率最高;结合CARS法对经FD法预处理后的光谱数据进行特征波长提取,共提取出22个波长;利用颜色矩和GLCM算法分别提取出每个牛肉样本的9个颜色特征、48个纹理特征。将特征波长数据与颜色、纹理特征信息进行融合建模,结果表明,基于特征光谱+纹理特征的模型识别效果最佳,其校正集与预测集识别率分别为98.42%和93.55%,均高于特征光谱数据模型识别率,说明融合纹理特征后使样本分类信息的表达更加全面;融合颜色特征后模型的校正集识别率均有所增加,但预测集识别率稍逊,颜色特征虽携带了部分有效信息,但这些信息与牛肉样本的相关性不大。因此,寻找与牛肉样本相关性更大的颜色特征是提高模型识别率的重要途径之一。该研究结果为牛肉品种的快速无损识别提供了一定的参考。     

Multi-sparse dictionary colorization algorithm based on the feature classification and detail enhancement

For the problems of missing details and performance of the colorization based on sparse representation, we propose a conceptual model framework for colorizing gray-scale images, and then a multi-sparse dictionary colorization algorithm based on the feature classification and detail enhancement (CEMDC) is proposed based on this framework. The algorithm can achieve a natural colorized effect for a gray-scale image, and it is consistent with the human vision. First, the algorithm establishes a multi-sparse dictionary classification colorization model. Then, to improve the accuracy rate of the classification, the corresponding local constraint algorithm is proposed. Finally, we propose a detail enhancement based on Laplacian Pyramid, which is effective in solving the problem of missing details and improving the speed of image colorization. In addition, the algorithm not only realizes the colorization of the visual gray-scale image, but also can be applied to the other areas, such as color transfer between color images, colorizing gray fusion images, and infrared images.