亚像素多重分形方法在图像处理中的应用

提出了一种用亚像素多重分形原理求取图像奇异性的新型算法,降低了单纯依靠整数像素位置灰度级梯度信息计算边缘测度所产生的误差。该算法结合CCD成像机理给出在亚像素位置的灰度级梯度分布规律,利用多重分形理论将实际图像分割成一系列具有不同奇异性指数的分形集合,对应着从边缘到纹理各层面的图像内容。模拟计算了投影小波中心点改变单位距离对边缘测度的影响程度,得出亚像素分割梯度的方法可以增加计算结果鲁棒性的结论。此方法用于标准图像的分割中,选用5×5亚像素数目提取的最奇异性集合与索贝尔(Sobel)算子(默认阈值为36.7920)提取的边缘的峰值信噪比为9.3981 dB。应用于复杂路面的裂纹提取中,其结果更符合人类的视觉观测。     

Wavelet-based multifractal analysis of DNA sequences by using chaos-game representation

Chaos game representation (CGR) is proposed as a scale-independent representation for DNA sequences and provides information about the statistical distribution of oligonucleotides in a DNA sequence. CGR images of DNA sequences represent some kinds of fractal patterns, but the common multifractal analysis based on the box counting method cannot deal with CGR images perfectly. Here, the wavelet transform modulus maxima (WTMM) method is applied to the multifractal analysis of CGR images. The results show that the scale-invariance range of CGR edge images can be extended to three orders of magnitude, and complete singularity spectra can be calculated. Spectrum parameters such as the singularity spectrum span are extracted to describe the statistical character of DNA sequences. Compared with the singularity spectrum span, exon sequences with a minimal spectrum span have the most uniform fractal structure. Also, the singularity spectrum parameters are related to oligonucleotide length, sequence component and species, thereby providing a method of studying the length polymorphism of repeat oligonucleotides.     

多重分形分析图象边缘提取算法

提出了一种基于多重分形分析的图象边缘提取算法,通过计算每个象素点的奇异值和多重分形谱,并根据多重分形谱的各种测度修正,提取出图象的边缘信息.在分析图象的各种象素点的多重分形谱特性的基础上,着重分析了多重分形谱常用的若干测度以及选取标准.该算法利用奇异性Hölder指数和多重分形谱以及它们组成的判据来进行图象边缘提取的思路不同于传统的基于梯度的局部极值点来进行图象边缘提取的方法.实验结果表明:该算法可以在保留重要边缘信息的情况下去除不重要细节,更能符合人的视觉心理.     

An improved computing method for the image edge detection

The framework of detecting the image edge based on the sub-pixel multi-fractal measure (SPMM) is presented. The measure is defined, which gives the sub-pixel local distribution of the image gradient. The more precise singularity exponent of every pixel can be obtained by performing the SPMM analysis on the image. Using the singularity exponents and the multi-fractal spectrum of the image, the image can be segmented into a series of sets with different singularity exponents, thus the image edge can be detected automatically and easily. The simulation results show that the SPMM has higher quality factor in the image edge detection.     

基于分类的红外云层背景弱小目标检测方法

提出了一种新的基于模糊分类的红外云层背景弱小目标检测方法.根据红外成像的特点,将红外云层背景弱小目标图像分为三类:边缘类、净空及云中类、弱小目标类;对不同类别图像进行分析,建立了分类模型,并定义了方向特征矢量,将其作为类别的特征矢量;根据模糊分类的理论,定义了类相似系数来判别图像中每一个像素的类别属性,保留弱小目标类的像素点完成检测.实验结果表明,该方法能够对红外弱小目标图像中不同类型的区域进行准确的分类,从而较好的实现了对低信杂比的复杂云层背景图像中的弱小目标检测.     

基于双树复小波变换的非平稳时间序列去趋势波动分析方法

多重分形去趋势波动分析是研究非平稳时间序列非均匀性和奇异性的有效工具, 针对该方法中趋势项难以确定的问题, 提出一种基于双树复小波变换的方法, 实现了非平稳信号的多重分形自适应去趋势波动分析. 利用双树复小波变换提取信号的多尺度趋势和波动信息, 通过小波系数的希尔伯特变换确定每个时间尺度不重叠子区间的长度, 使多重分形分析具有信号自适应性及较高的计算效率. 以具有解析形式分形特征的倍增级联信号和分数布朗运动时间序列为例验证本文方法的有效性, 所得结果与解析解相吻合. 与传统的多项式去趋势多重分形方法相比, 本文方法根据信号自身特点自适应地确定信号的趋势和不重叠等长度子区间长度, 所得结果更加精确. 对倍增级联信号时间序列取不同的长度, 验证了算法的稳定性. 分别与基于极大重叠离散小波变换和离散小波变换多重分形方法进行比较, 表明本文方法具有更精确的结果和更快的运算速度.     

基于自适应扩散模型的单帧红外条纹非均匀性校正算法

针对红外焦平面成像系统存在列向条纹非均匀性的现象,采用了一种基于自适应PM扩散模型的非均匀校正新算法。首先,综合利用图像梯度信息和局部灰度统计信息,自适应计算PM模型的扩散阈值;然后将每列像素的PM模型估计值作为该列像素的期望值;最后采用最陡下降法迭代计算得到每列像元的校正参数,并对结果进行循环校正以提高校正效果。实验结果表明:该算法可以保护图像边缘信息,与同类算法相比,能够更有效地抑制条纹非均匀性,并且能够防止图像产生鬼影。     

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.     

潜在低秩表示框架下基于卷积神经网络结合引导滤波的红外与可见光图像融合

为提高融合图像的可视性,解决传统红外与可见光图像融合算法中存在的边缘特征缺失、细节模糊的问题,提出了一种潜在低秩表示框架下基于卷积神经网络结合引导滤波的红外与可见光图像融合算法。该算法首先利用潜在低秩表示对源图像进行分解,得到源图像的低秩分量和显著分量。其次,利用卷积神经网络根据源图像的特征信息,得到权值图。再次,通过引导滤波算法对权值图进行边缘锐化,然后再将优化后的权值图分别与源图像的低秩分量和显著分量融合,得到融合图像的低秩分量和显著分量。最后,将融合图像的低秩分量和显著分量叠加,得到最终的融合图像。实验结果表明,该算法在主观评价和客观指标上均优于传统的红外与可见光图像融合算法。     

基于红外图像边缘特性的自动对焦技术

与可见光图像相比,红外图像的分辨率较低,景物边缘较为粗糙,传统的自动调焦技术无法直接用于红外热成像系统。为了解决这一问题,采用基于红外图像边缘特性的自动对焦算法,建立快速有效的对焦评价函数,驱动高精度光机闭环控制单元,以实现快速、准确的红外自动对焦。系统包含数字处理单元和运动控制单元。其中数字图像处理单元主要进行红外数字图像的采集,利用Roberts梯度算子提取红外图像中的信息量,并统计边缘点数以评价图像的清晰度,实时计算序列红外图像中细节和信息量的变化量和趋势,从而提供运动控制单元所需的位置给定。运动控制单元则主要进行调焦电机的控制及驱动,完成自动聚焦的过程控制。最后,通过实验证明了该方案的有效性。     

基于小波leaders的海杂波时变奇异谱分布分析

海杂波的奇异谱分析不仅能从理论上揭示海洋表面的动力学机理,同时也是对海探测雷达的关键技术之一.本文提出基于小波leaders的海杂波时变奇异谱分析方法,将时间信息引入海杂波的奇异谱分析之中,从而实现动态的解析描述海杂波随时间变化的奇异谱特性.在理论上,通过信号自身加窗,将时间信息引入传统的奇异谱(或称多重分形谱),实现了对海杂波时变奇异谱分布分析;在算法上,充分利用了小波leaders技术对于多种奇异性的提取能力(包括chirp奇异性和cusp奇异性),通过对时变奇异性指数和时变尺度函数的Legendre变换,实现对海杂波时变奇异谱分布的计算;在应用部分,采用经典的多重分形模型——随机小波序列(RWC)以及三级海态条件下连续波多普勒体制雷达海杂波进行仿真分析,实验结果表明:1)基于小波leaders的奇异谱分布能跟踪海杂波的时变尺度特性,有效展示其时变奇异性谱分布;2)算法具有较好的负矩特性和统计收敛性.该方法能为复杂非线性系统及随机多重分形信号分析提供参考.     

Improving spatio-temporal resolution of infrared images to detect thermal activity of defect at the surface of inorganic glass

This paper proposes a noise suppression methodology to improve the spatio-temporal resolution of infrared images. The methodology is divided in two steps. The first one consists in removing the noise from the temporal signal at each pixel. Three basic temporal filters are considered for this purpose: average filter, cost function minimization (FIT) and short time Fast Fourier Transform approach (STFFT). But while this step effectively reduces the temporal signal noise at each pixel, the infrared images may still appear noisy. This is due to a random distribution of a residual offset value of pixels signal. Hence in the second step, the residual offset is identified by considering thermal images for which no mechanical loading is applied. In this case, the temperature variation field is homogeneous and the value of temperature variation at each pixel is theoretically equal to zero. The method is first tested on synthetic images built from infrared computer-generated images combined with experimental noise. The results demonstrate that this approach permits to keep the spatial resolution of infrared images equal to 1 pixel. The methodology is then applied to characterize thermal activity of a defect at the surface of inorganic glass submitted to cyclic mechanical loading. The three basic temporal filters are quantitatively compared and contrasted. Results obtained demonstrate that, contrarily to a basic spatio-temporal approach, the denoising method proposed is suitable to characterize low thermal activity combined to strong spatial gradients induced by cyclic heterogeneous deformations.     

基于深度优先遍历的图像边缘检测方法

图像边缘识别是图像处理的重要组成。提出一种基于深度优先遍历的梯度分割算法,这种算法首先构造像素点的数据结构,然后从图像任意点出发,估算该点附近的像素点并获取梯度值,如果该点满足边缘点的特征,则从该点出发深度遍历寻找垂直于梯度方向上的边缘点,并标记访问过的点。如果该方向上没有满足条件的点则回退,从某个具有仅次于最大梯度值的方向继续遍历,并标记开始遍历的点为角点,直至遍历全图。该算法将图像的边缘点和角点明显地分割出来,便于识别,对带有不同类型的噪声图像进行处理也可取得较满意效果。     

基于支持度变换和top-hat分解的双色中波红外图像融合

为了解决用多尺度top-hat分解法融合双色中波红外图像时经常存在对比度提升有限、边缘区域失真较重的问题,提出了基于支持度变换和top-hat分解相结合的融合方法。先用支持度变换法将双色中波图像分解为低频图像和支持度图像序列;再从最后一层低频图像中用多尺度top-hat分解法提取各自的亮信息和暗信息;用灰度值取大法分别融合亮信息和暗信息;通过灰度值归一化和高斯滤波分别增强亮、暗信息融合图像;然后融合两低频图像和亮、暗信息增强图像;将融合图像作为新的低频图像和用灰度值取大法融合得到的支持度融合图像序列进行支持度逆变换,得到最终融合图像。该方法的实验结果同采用单一的支持度变换法融合和多尺度top-hat分解法融合相比,融合图像的对比度提升了11.69%,失真度降低了63.42%,局部粗糙度提高了38.12%。说明提出的从低频图像提取亮暗信息,并经过分别融合、增强,再与低频图像进行融合,能有效破解红外融合图像对比度提升和边缘区域失真度降低之间的矛盾,为提高图像融合质量提供了新方法。     

基于多任务卷积神经网络的红外与可见光多分辨率图像融合

红外与可见光图像融合一直是图像领域研究的热点,融合技术能弥补单一传感器的不足,为图像理解与分析提供良好的成像基础。因生产工艺以及成本的限制,红外探测器的分辨率远低于可见光探测器,并在一定程度上因源图像分辨率的差异阻碍了实际应用。针对红外与可见光图像分辨率不一致的问题,提出了用于红外图像超分辨率重建与融合的多任务卷积网络框架,应用于多分辨率图像融合。在网络结构方面,首先设计了双通道网络分别提取红外与可见光特征,使算法不受源图像分辨率的限制;其次提出了特征上采样模块,先用双线性插值方法增加像素个数,再通过多层感知器精细化拟合像素平滑空间与高频空间的映射关系,无需重新训练模型即可实现任意尺度的红外图像上采样;接着将线性注意力引入网络,学习特征空间位置间的非线性关系,抑制无关信息并增强网络对全局信息的表达。在损失函数方面,提出了梯度损失,保留红外与可见光图像中绝对值较大的滤波器响应值,并计算该值与重建的融合图像响应值的Frobenius范数,无需理想的融合图像作为真值监督网络学习就能生成融合图像;此外,在梯度损失、像素损失的共同作用下对多任务模型进行优化,可以同时重建融合图像和高分辨率红外图像...     

一种基于小波变换的红外偏振融合算法

针对红外偏振图像可以较好地抑制背景噪声,对目标边缘信息比较敏感的特点,提出一种基于小波变换的红外偏振融合算法,它主要用于红外辐射强度图像和偏振度图像融合,增加图像的信息量。首先采用小波变换对参与融合的每幅图像分别进行各尺度分解,得到各尺度小波系数,然后针对不同尺度小波系数,采用邻域平均梯度为判据进行融合,得到融合后的各尺度小波系数,最后通过小波逆变换进行图像重构,得到融合图像。融合前后的图像对比表明融合图像在保留辐射强度图像的清晰度的同时,突出了目标的边缘、轮廓信息。相对于辐射强度图像,融合图像的梯度均值提高了112%,相对于偏振度图像,融合图像的标准差提高了151%,信息熵提高了38%。     

高光谱成像技术无损检测赣南脐橙表面农药残留研究

高光谱成像技术具备图像和光谱的双重优势,作为一种快速无损检测分析技术,检测过程无损、无污染和无接触。高光谱成像数据包括样本的图像信息和光谱信息,采集样本高光谱成像数据时,样本的每个像素点都有一条光谱与之对应,样本的每个波长都有一幅灰度图像与之对应。研究采用高光谱成像技术无损检测不同稀释浓度的农药在赣南脐橙样品表面残留随时间变化的关系。用蒸馏水把农药分别配置成1∶20, 1∶100和1∶1 000倍的溶液。然后把不同浓度的溶液滴到30个洗净的脐橙表面, 将涂有农药的脐橙分别放置0,4和20 d,然后采集在900～1 700 nm波长范围的高光谱成像原始数据。通过主成分分析获取930,980,1 100,1 210,1 300,1 400,1 620和1 680 nm共8个特征波长,基于这些特征波长做第二次主成分分析,应用PC-2图像并经过适当的图像处理方法对不同浓度及放置不同天数的农药残留进行无损检测。采用高光谱成像技术检测三个时间段较高稀释浓度的果面农药残留都比较明显。高光谱成像技术作为一种检测方法,可用于评价各个时间段较高浓度的农药残留。     

Detection of meso-micro scale surface features based on microcanonical multifractal formalism

Synthetic aperture radar(SAR) is an effective tool to analyze the features of the ocean. In this paper, the microcanonical multifractal formalism is used to analyze SAR images to obtain meso-micro scale surface features. We use the Sobel operator to calculate the gradient of each point in the image, then operate continuous variable scale wavelet transform on this gradient matrix. The relationship between the wavelet coefficient and scale is built by linear regression. This relationship decides the singular exponents of every point in the image which contain local and global features. The manifolds in the ocean can be revealed by analyzing these exponents. The most singular manifold, which is related to the streamlines in the SAR images, can be obtained with a suitable threshold of the singular exponents. Experiments verify that application of the microcanonical multifractal formalism to SAR image analysis is effective for detecting the meso-micro scale surface information.     

Large deviations estimates for the multiscale analysis of heart rate variability

In the realm of multiscale signal analysis, multifractal analysis provides a natural and rich framework to measure the roughness of a time series. As such, it has drawn special attention of both mathematicians and practitioners, and led them to characterize relevant physiological factors impacting the heart rate variability. Notwithstanding these considerable progresses, multifractal analysis almost exclusively developed around the concept of Legendre singularity spectrum, for which efficient and elaborate estimators exist, but which are structurally blind to subtle features like non-concavity or, to a certain extent, non scaling of the distributions. Large deviations theory allows bypassing these limitations but it is only very recently that performing estimators were proposed to reliably compute the corresponding large deviations singularity spectrum. In this article, we illustrate the relevance of this approach, on both theoretical objects and on human heart rate signals from the Physionet public database. As conjectured, we verify that large deviations principles reveal significant information that otherwise remains hidden with classical approaches, and which can be reminiscent of some physiological characteristics. In particular we quantify the presence/absence of scale invariance of RR signals.     

基于边缘与SURF算子的SAR与可见光图像配准方法

鉴于SAR(synthetic aperture radar)与可见光图像的成像机理存在很大差别,使得其同名特征的提取和配准十分困难,但在某些情况下,这两类图像的边缘存在一定的相关性。提出一种基于边缘与SURF(speed-up robust feature)算子的图像配准方法。通过适当预处理增强图像间的共性,采用综合性能比较好的Canny算子提取两幅图像共有的边缘特征,在边缘图像的基础上提取SURF特征;通过比值提纯法进行特征点粗匹配,RANSAC(random sample consensus)算法剔除误匹配点,计算仿射变换模型从而实现SAR与可见光图像的自动配准。实验结果表明:该算法的正确匹配率为100%,均方根误差为0.852个像素,配准精度达到亚像素水平。