基于同色异谱黑理论的非线性复合模型光谱重建研究

同色异谱现象是光谱反射率重建与颜色再现中的一个重要问题。采用三基色CCD相机采集CIE标准光源D65下的颜色样品信号,建立非线性复合模型,使用主成分分析结合神经网络的方法(PCA-NET)改进基于同色异谱黑理论的R矩阵算法,对标准Munsell色卡光谱重建进行研究。在保证给定照明条件下的色度精度同时,对光谱重建的结果进行了实验评价和讨论。实验结果表明,在给定的照明条件下,PCA-NET算法能够准确的拟合相机输出信号与主成分系数之间的非线性关系, 将其代替线性算法应用于R矩阵算法中时,测试集的平均均方根值是未改进R矩阵算法的0.76,平均标准差是R矩阵算法的0.85,可有效提高光谱反射率的重建精度。改进后的R矩阵算法具有精度较高、操作简单易实现的特点, 可用于对重建色度精度及光谱精度均要求较高的领域。     

动态混沌扰动遗传算法的振动模糊高光谱图像复原

高光谱图像具有较高的光谱分辨率和空间分辨率,从而具备区分诊断地物光谱特性的能力,但是在获取高光谱图像时,经常会由于载荷平台的振动,导致光谱图像失真,严重影响光谱图像在应用中的精度和可信度。提出一种能够将振动模糊高光谱图像复原的动态混沌扰动遗传算法, 该算法对比于普通的遗传算法,不会出现过早收敛,能够较准确的恢复图像,提高光谱质量。根据振动模糊图像的退化原理,找到振动模糊图像与清晰图像之间的映射关系和振动模糊图像的点扩散函数。针对振动模糊图像退化的非线性和混沌系统特征,用tent映射生成混沌初始种群,增强遗传算法的全局搜索能力。对产生的优秀个体,用切比雪夫映射进行混沌扰动,对优秀个体混沌优化,以增强遗传算法自身局部搜索能力。将三维高光谱影像平铺为二维图像,利用相邻的光谱通道的图像相关性,对每一幅图像进行复原,从而实现三维高光谱数据的复原。在澳大利亚机载Hymap成像光谱仪所提供的数据立方体中,进行了两组不同的振动模糊光谱图像复原仿真验证。将所提出的方法与近期使用的光谱图像复原算法和遗传复原算法进行对比分析,图像采用无参评价方法灰度平均梯度GMG和拉普拉斯算子LS,有参评价方法信噪比SNR和峰值信噪比PSNR,光谱采用光谱信息散度SID和光谱梯度角SGA评价方法,发现各个评价指标均有大幅改善。与最新的光谱复原算法相比,SNR提高了60%,PSNR提高了10%,GMG提高了11%,LS提高了11%,SID降低了39%,SGA降低了5%。与原遗传算法相比,图像的SNR提高了51%,PSNR提高了12%,GMG提高了33%,LS提高了43%,SID降低了39%,SGA降低了16%。计算结果表明该方法对恢复振动模糊高光谱图像数据非常有效,不仅能提高单波段图像的清晰度,光谱数据立方体的光谱质量也明显提高。     

PCB光电图像光照不均蜕化的校正及阈值分割方法

分析了CCD获取的图像光照不均蜕化产生原因.针对光照不均蜕化的印刷电路板图像的具体情况,用同态滤波器在频域空间进行滤波而增强图像,从而改善图像的目标与背景对比度.讨论了最大类间方差法,给出了其优点.为了更好的分割图像,在此基础上,考虑两类间距与各类内聚性对图像分割的不同影响,提出了一种改进的最大类间方差法.最后对光照不均蜕化的PCB光电图像用不同方法进行了实验,结果表明改进方法能更好地分割图像.     

光电图像光照不均蜕化的校正及阈值分割方法*

分析了CCD获取的图像光照不均蜕化产生原因.针对光照不均蜕化的印刷电路板图像的具体情况,用同态滤波器在频域空间进行滤波而增强图像,从而改善图像的目标与背景对比度.讨论了最大类间方差法,给出了其优点.为了更好的分割图像,在此基础上,考虑两类间距与各类内聚性对图像分割的不同影响,提出了一种改进的最大类间方差法.最后对光照不均蜕化的PCB光电图像用不同方法进行了实验,结果表明改进方法能更好地分割图像.     

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.     

Partial Exclusion of the Local Effect from the Assessment of Recognized Illuminant Using Depth Separation II: Influence of Depth Separation under Chromatic Illumination

In the previous work we investigated the influence of depth separation on exclusion of the local effect from the assessment of the recognized illuminant. Under whitish illumination, only a portion of the local effect was excluded, while this study, we extended the range of illumination from whitish to chromatic illuminant. We evaluated how the depth separation influences the amount of local effect under the chromatic illumination. Subjects did perceptual achromatic setting under various combinations of local surround and illumination conditions. We found that under chromatic illumination, depth separation is more effective to exclude the local effect even though the local effect was smaller than under whitish chromatic illumination.     

基于旋转SURF算子的图像配准新方法

针对SURF算法中快速Hessian矩阵行列式检测出的特征点的不连续现象,从而造成的旋转,模糊和光照变化适应性较差的不足,提出一种旋转SURF检测算子的图像配准新方法。该算法通过将SURF算法的积分图像盒子滤波模板逆时针旋转45度,引入一种可以检测角度旋转的滤波核提升检测算子对不同图像变换的匹配性能,保证新的检测算子与原算法较好的结合,同时利用改进的单纯形算法依据输入图像进行参数优化。仿真结果表明,该方法不仅保留了算法的速度优势,缩短了配准时间,而且在图像模糊变换,光照变换和JPEG压缩变换方面性能有明显的提升,此外对视角变换以及小尺度变换性能也有提高。     

基于自适应阈值方法实现迭代降噪鬼成像

为了有效降低传统鬼成像中相关噪声对成像质量的影响,本文提出一种基于最佳阈值的迭代降噪鬼成像.首先在迭代降噪鬼成像的基础上,采用自适应阈值迭代法,在不需要目标先验信息的前提下,找到一个逼近传统鬼成像中相关噪声的阈值,根据得到的阈值构造噪声干扰项.为了每次迭代初值更接近原始目标的透射系数,对其进行二值化,降低重构图像背景噪声对迭代性能的影响.仿真以及实验结果表明,本文提出的方法与传统方法相比,视觉效果以及峰值信噪比值有明显提高.     

基于参数展开的同伦分析法在强非线性随机动力系统中的应用

将基于参数展开的同伦分析法(PE-HAM)进行了推广，使之适用于谐和激励与随机噪声联合作用下的强非线性随机动力系统. 通过构造合适的同伦映射，将对强非线性随机动力系统响应的求解转化为对一组线性随机微分方程的求解. 进一步研究了受到谐和与Gauss白噪声激励的强非线性Duffing振子，由PE-HAM得到了该系统的解过程和稳态概率密度的解析表达式. 数值模拟的结果说明了PE-HAM方法的精确性. 关键词： PE-HAM方法 强非线性随机动力系统 稳态概率密度 解过程 随机激励     

彩色图射中颜色不变性和耀斑的分离(Ⅱ)

光源的位置和色品影响景物在摄像机中的成像,耀斑和阴影的出现使一般的分割算法得不到理想的分割结果。本文根据双色反射模型将图像分解成两个本征图像——本色图像(matte image)和耀斑图像(highlighit image)。本色图像是去掉耀斑影响的图像,反映了景物本身的不变性,不受外部因素(光源的位置和色品)的影响。而耀斑图像是只含有耀斑的图像,恰反映了外部因素的影响。并提供景物形状的信息。     

Finger crease pattern recognition using Legendre moments and principal component analysis

The finger joint lines defined as finger creases and its distribution can identify a person. In this paper,we propose a new finger crease pattern recognition method based on Legendre moments and principal component analysis (PCA). After obtaining the region of interest (ROI) for each finger image in the preprocessing stage, Legendre moments under Radon transform are applied to construct a moment feature matrix from the ROI, which greatly decreases the dimensionality of ROI and can represent principal components of the finger creases quite well. Then, an approach to finger crease pattern recognition is designed based on Karhunen-Loeve (K-L) transform. The method applies PCA to a moment feature matrix rather than the original image matrix to achieve the feature vector. The proposed method has been tested on a database of 824 images from 103 individuals using the nearest neighbor classifier. The accuracy up to 98.584% has been obtained when using 4 samples per class for training. The experimental results demonstrate that our proposed approach is feasible and effective in biometrics.     

基于光谱线性拟合的宽波段光源波动补偿方法

宽波段光源在工作过程中,光强通常会随供电电源输出功率的变化而波动。宽波段光源波段内各个波长的光谱强度将会发生不同程度的波动。为解决光源光强波动时其波段内各个波长光谱强度的补偿问题,提出了一种基于光谱线性拟合的补偿方法。使用这种方法,只需测量光源波段光强的变化,就可以补偿各个波长光谱强度的波动。通过分析理想黑体全波段辐射出射度与光谱辐射出射度的近似线性关系。建立了宽波段光源波段光强与光谱强度的线性模型。实验装置主要由卤素灯珠、光源电源、光阑、光谱仪及计算机构成。调节电源的输出功率,得到一组卤素灯珠在不同输入功率下的相对光谱强度。测量不同功率下卤素灯珠光谱强度来验证该方法补偿效果。按线性关系拟合了卤素灯珠光谱强度与其波段光强关系式,并分析了拟合误差。实验表明：卤素灯珠的光谱强度与其波段光强具有很好的线性关系,可以用波段光强按线性关系来补偿其光谱强度的波动。随着卤素灯珠输入功率的增大,补偿后的光谱强度的相对误差绝对值下降。在卤素灯珠输入功率范围内,使用该方法补偿的光谱强度在绝大部分(92%)波长下相对误差绝对值可在5%以内。     

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.     

Kernel-aligned multi-view canonical correlation analysis for image recognition

Existing kernel-based correlation analysis methods mainly adopt a single kernel in each view. However, only a single kernel is usually insufficient to characterize nonlinear distribution information of a view. To solve the problem, we transform each original feature vector into a 2-dimensional feature matrix by means of kernel alignment, and then propose a novel kernel-aligned multi-view canonical correlation analysis (KAMCCA) method on the basis of the feature matrices. Our proposed method can simultaneously employ multiple kernels to better capture the nonlinear distribution information of each view, so that correlation features learned by KAMCCA can have well discriminating power in real-world image recognition. Extensive experiments are designed on five real-world image datasets, including NIR face images, thermal face images, visible face images, handwritten digit images, and object images. Promising experimental results on the datasets have manifested the effectiveness of our proposed method.     

An improved hierarchical fragile watermarking scheme using chaotic sequence sorting and subblock post-processing

The original hierarchical watermarking scheme for image tamper detection and recovery has simple computation and high performance precision which can achieve 2 × 2 subblock. However, four-scanning and blind attacks have been proposed recently on this scheme. We generalize these attacks and analyze the cause of security flaws. We think that it is promising to improve the original scheme's security and keep its merit at the same time. In order to defeat these attacks, we develop an improved method to generate a block mapping sequence by sorting the chaotic sequence and add the chaotic encryption and permutation based on the exact content of each subblock to be the post-processing of the 3-tuple watermark. Our method uses the simple watermarking scheme and satisfies the performance requirements of fragile watermarking, such as high-precision tamper detection, localization and recovery. Theoretical analysis and computer simulation demonstrate that our proposed scheme is much more secure and can overcome possible attacks.     

Effect of coherence and polarization on resolution of optical imaging system

The effect of coherence properties of illumination on image resolution, well known in a scalar case, is studied for the case of vector electromagnetic illumination. Using an example of vector Gaussian Schell-model illumination, we analyze the dependence of optical system resolution on the transverse correlation lengths of the orthogonal field components and on the ratio of the powers of these components, each taken separately.     

Diffraction microtomography with sample rotation: influence of a missing apple core in the recorded frequency space

Diffraction microtomography in coherent light is foreseen as a promising technique to image transparent living samples in three dimensions without staining. Contrary to conventional microscopy with incoherent light, which gives morphological information only, diffraction microtomography makes it possible to obtain the complex optical refractive index of the observed sample by mapping a three-dimensional support in the spatial frequency domain. The technique can be implemented in two configurations, namely, by varying the sample illumination with a fixed sample or by rotating the sample using a fixed illumination. In the literature, only the former method was described in detail. In this report, we precisely derive the three-dimensional frequency support that can be mapped by the sample rotation configuration. We found that, within the first-order Born approximation, the volume of the frequency domain that can be mapped exhibits a missing part, the shape of which resembles that of an apple core. The projection of the diffracted waves in the frequency space onto the set of sphere caps covered by the sample rotation does not allow for a complete mapping of the frequency along the axis of rotation due to the finite radius of the sphere caps. We present simulations of the effects of this missing information on the reconstruction of ideal objects.      

Personalized Image Classification by Semantic Embedding and Active Learning

Currently, deep learning has shown state-of-the-art performance in image classification with pre-defined taxonomy. However, in a more real-world scenario, different users usually have different classification intents given an image collection. To satisfactorily personalize the requirement, we propose an interactive image classification system with an offline representation learning stage and an online classification stage. During the offline stage, we learn a deep model to extract the feature with higher flexibility and scalability for different users’ preferences. Instead of training the model only with the inter-class discrimination, we also encode the similarity between the semantic-embedding vectors of the category labels into the model. This makes the extracted feature adapt to multiple taxonomies with different granularities. During the online session, an annotation task iteratively alternates with a high-throughput verification task. When performing the verification task, the users are only required to indicate the incorrect prediction without giving the exact category label. For each iteration, our system chooses the images to be annotated or verified based on interactive efficiency optimization. To provide a high interactive rate, a unified active learning algorithm is used to search the optimal annotation and verification set by minimizing the expected time cost. After interactive annotation and verification, the new classified images are used to train a customized classifier online, which reflects the user-adaptive intent of categorization. The learned classifier is then used for subsequent annotation and verification tasks. Experimental results under several public image datasets show that our method outperforms existing methods.     

Three-dimensional image correlator using fast computational integral imaging reconstruction method based on pixel-to-pixel mapping

In this paper, a three-dimensional (3D) image correlator using a fast computational integral imaging reconstruction (CIIR) method based on a pixel-to-pixel mapping is proposed. In order to implement the fast CIIR method, we replace the magnification process in the conventional CIIR by a pixel-to-pixel mapping. The proposed fast CIIR method reconstructs two sorts of plane images; a plane image whose quality is sufficient, and a dot pattern plane image insufficient to view. This property is very useful to enhance the performance of a CIIR-based image correlator. Thus, we apply the fast CIIR method to a CIIR-based image correlator. To show the feasibility of the proposed method, some preliminary experiments on both pattern correlation and computational cost are carried out, and the results are presented. Our experimental results indicate that the proposed image correlator is superior to the previous method in terms of both correlation performance and complexity.