基于结构-纹理分解的数字图像分层修复算法

采用目前方法进行数字图像修复时,没有对边缘的衔接效果和灰度跳变问题进行考虑,导致视觉效果差、SSIM与PSNR值低的问题。提出基于结构-纹理分解的数字图像分层修复算法,首先运用TV-L~1模型完成数字图像的结构-纹理拆分,然后利用快速行进法对图像进行预填充,最后对数字图像结构-纹理部分进行分层修复,在结构部分运用结合12个邻域点的加权组合更新受损点,获取更好的衔接效果,纹理部分采用Criminisi算法,用局部搜索代替全局搜索,完成数字图像的分层修复。通过实验结果可以看出,所提方法能够有效地改善视觉效果,提高SSIM与PSNR值。

基于图像结构-纹理分解及局部总变分最小化的图像修复模型

在基于样例的图像修复算法中,由于优先权公式的计算容易受图像局部噪声和细小纹理的干扰,导致修复顺序错乱;而在搜索最优匹配块时,因忽略了图像块内部的结构影响,可能导致误匹配。针对以上问题提出了一种基于图像的结构-纹理分解及局部总变分最小化的图像修复模型。首先,根据对数总变分最小化模型,将待修复图像进行结构-纹理分解,得到图像的结构分量,并利用图像的结构分量来计算待修复点优先权,使优先权的计算排除局部纹理干扰而更具鲁棒性;其次,将优先权的计算改进为数据项和置信项的加权和,避免了乘积效应,确保数据项一直发挥作用,减少因修复顺序不合理造成的错误匹配;最后,根据图像的局部总变分最小化原则,将图像块的最优匹配转换为0-1优化问题,确保图像修复后的局部结构一致性。与3组参考文献的5组对比实验结果表明,峰值信噪比(PSNR)提高了1.12~3.56 d B,结构相似性指数提高了0.02~0.04。所提模型更好地遵循了修复优先性原则,具有更强的保持图像局部结构一致性的能力,改善了修复图像的视觉效果,适用于复杂结构的大面积毁损的图像的修复。

结构-纹理字典学习的图像分解模型与算法

为了分离出图像中具有不同特征的成分,结合变分与字典学习方法,提出一种图像分解模型和结构-纹理字典学习算法.首先在模型中引入字典约束项,使得结构-纹理学习字典互不相关,增强了2个字典的独立性;然后使用投影梯度下降算法给出一种带有字典约束的交替字典学习算法.实验结果表明,采用该算法学习得到的自适应字典可以有效地刻画图像的不同成分,不仅很好地分开了图像的结构和纹理,并且能去除噪声,最终得到高质量的图像分解结果.

基于结构-纹理分解和VGG深层网络的红外与可见光图像融合

针对当前红外与可见光图像融合方法中存在低频信息未充分利用,以及高频细节易混入噪声等问题,提出一种基于结构-纹理(ST)分解和VGG深层网络的红外与可见光图像融合方法。首先用均值滤波将输入图像分解为高-低频子带,并引入ST对低频子带二次分解,采用平均法则与邻域空间频率分别对结构和纹理进行预融合;同时,将输入图像送入VGG网络得到多层特征映射,并用Sigmiod函数实现高频子带的归一化预融合;最后利用预融合的高频、低频结构和低频纹理3个频带进行图像融合重建。实验结果表明,提出的方法能够融合图像的深层细节特征,有效保留纹理细节并抑制噪声,且在噪声评估、结构相似性、均方误差、峰值信噪比等客观指标方面具有明显优势。

基于Mumford-Shah模型和G空间的图像结构纹理分解

构造了一个变分问题来实现图像的纹理结构分解。其中结构成分用分段光滑的函数(即Mumford-Shah模型)刻画,纹理部分用振荡函数(G空间)来描述。由于Mumford-Shah模型将结构成分显式地描述为分段光滑函数,在非结构边缘点处的梯度采用L2范数约束,故相对于全变差(Total variation,TV,梯度的L1范数)的分解方式,很好地克服TV带来的阶梯效应;G空间本身定义的纹理函数的振荡特性保证了分解的结构成分中含有更少的纹理信息。实验表明,无论相对于经典的TV-L2分解还是TV-G分解,本文方法均体现出了很好的分解性能。

基于结构-纹理分层的夜间图像去雾算法

夜间图像光照不均匀,存在色偏,去雾难度较大。目前图像去雾算法主要针对白天场景,有关夜间图像去雾算法的研究较少。基于结构-纹理分层模型提出新的夜间图像去雾算法,将夜间有雾图像分解为结构层和纹理层。在结构层采用中值滤波器估计环境光,利用加权范数L1正则化模型对其进行优化,并进行去雾和颜色校正处理;在纹理层利用离散余弦变换系数估计透射率。最终融合纹理层与去雾后的结构层得到去雾图像。实验结果表明,采用该算法对夜间图像去雾后图像细节清晰,颜色自然,去雾效果显著。

基于结构纹理分解的改进图像压缩算法

用等级树集合分区(set partitioning in hierarchical trees,SPIHT)压缩结构分量所产生的误差一般情况下能体现高振荡的特性,比较符合纹理特点,所以将这部分误差信号叠加在纹理分量中形成一个新的纹理源,利用小波包变换进行压缩,提出了一种改进的基于结构-纹理分解的图像压缩处理算法.理论分析表明这是分层压缩问题中一种普适的有效策略.从实验结果可以看出,这种处理方式相对于现有的结构纹理分解压缩的方式而言,始终具有一定的信噪比增益,相对于目前的JPEG2000压缩标准及单一的SPIHT、快速小波包变换(FWP)等算法具有一定的优越性.

联合语义边缘与有向全变分的纹理-结构分解

在去除尺度变化较大的纹理时,基于低级视觉特征的纹理-结构分解方法难以准确保留语义上重要的弱边缘,针对该问题提出了融合语义边缘检测与有向全变分模型的纹理-结构分解方法.首先在不同尺度下利用多卷积特征网络估计各像素的语义边缘置信度;然后在纹理的局部振荡假设基础上以有向全变分估计各像素的纹理度,为进一步有效地区分纹理边缘和结构边缘,引入块平移算法修正纹理度,并结合边缘置信度优化分解模型的权重.在BSDS500, NYUD以及RTVD数据集上的实验结果表明,该方法在分解准确性和视觉质量上优于现有多种方法,且易于通过GPU加速高分辨率图像的分解效率.

纹理细节恢复的图像超分辨率重建算法

针对现有单图像超分辨率方法在重建过程中容易忽略原图像中不同结构-纹理的差异与联系,导致生成的高分辨率图像缺乏纹理细节并存在伪影的问题,提出了纹理细节恢复的图像超分辨率重建算法。该方法由梯度分支、纹理分支和图像超分辨率分支组成。其中,在梯度分支和纹理分支之间使用了类注意力模块处理二者的特征混淆问题,并通过双向特征融合模块实现了对结构特征与纹理特征的相互促进,作为先验信息以达到纹理细节信息增强的目的。此外,在图像超分辨率分支还通过构建特征恢复模块,利用浅层和深层信息帮助网络保留了图像中更丰富的上下文信息和纹理细节。该方法通过在DIV2K数据集上进行了网络训练,并在5个基准测试集Set5、Set14、BSD100、Urban100和MANGA109上进行了实验,峰值信噪比(PSNR, Peak Signal to Noise Ratio)分别:37.88dB、33.28dB、32.0781dB、31.89dB、38.39dB,相比现有方法均有显著提升。实验结果表明,本文方法获得了有效的重建图像并且保留更多的图像细节,生成具有边缘清晰和逼真细节的超分辨率图像。

Texture features analysis on micro-structure of paste backfill based on image analysis technology

The strength of cement-based materials,such as mortar,concrete and cement paste backfill(CPB),depends on its microstructures(e.g.pore structure and arrangement of particles and skeleton).Numerous studies on the relationship between strength and pore structure(e.g.,pore size and its distribution)were performed,but the micro-morphology characteristics have been rarely concerned.Texture describing the surface properties of the sample is a global feature,which is an effective way to quantify the micro-morphological properties.In statistical analysis,GLCM features and Tamura texture are the most representative methods for characterizing the texture features.The mechanical strength and section image of the backfill sample prepared from three different solid concentrations of paste were obtained by uniaxial compressive strength test and scanning electron microscope,respectively.The texture features of different SEM images were calculated based on image analysis technology,and then the correlation between these parameters and the strength was analyzed.It was proved that the method is effective in the quantitative analysis on the micro-morphology characteristics of CPB.There is a significant correlation between the texture features and the unconfined compressive strength,and the prediction of strength is feasible using texture parameters of the CPB microstructure.

基于相对总变差模型与自适应形态学的织物瑕疵检测

为解决目前基于图像处理的织物瑕疵检测算法中,因织物纹理的多样性与瑕疵形状尺寸的不确定性所造成的检测效果差的问题,提出一种基于结构-纹理模型与自适应数学形态学的织物瑕疵检测算法。首先采用相对总变差模型对织物图像进行滤波以去除织物纹理,然后在得到的灰度图像上直接进行基于自适应邻域的灰度形态学运算,形态学算子采用开运算算子,最终得到织物瑕疵的增强图像。采用基于相对总变差模型与自适应形态学相结合的方法与2种已知的Gabor算法进行比对,对4类典型织物瑕疵进行检测实验和分析。结果表明,本文方法能更好地提取织物瑕疵。

Dual optimization image repair algorithm based on linear structure and optimal texture

The performances of repaired image depend on the local information in the repaired area and the consistency between the repair directions with structural content.Image repair algorithm with texture information performs well in repairing seriously damaged images,but it has bad performances when the images have the abundant structure information.The dual optimization image repair algorithm based on the linear structure and the optimal texture is proposed.The algorithm uses the double-constraint sparse model to reconstruct the missed information in large area in order to improve the clarity of repaired images.After adopting the preference of Criminisi priority,the image repair algorithm of self-similarity characteristics is proposed to improve the fault and fuzzy distortion phenomena in the repaired image.The results show that the proposed algorithm has more clarity in the image texture and structure and better effectiveness,and the peak signal-to-noise ratio of the repaired images by proposed algorithm is superior to that by other algorithms.

Effects of microstructure and texture after thermomechanical treatments on corrosion behavior of AISI 321 pipeline austenitic stainless steel

In the present study,the effects of microstructure,grain size,and texture after thermomechanical processing on the corrosion behavior of AISI 321 austenitic stainless steel(ASS)were studied.The as-received,coarse-grained steel((35±3)μm)was subjected to 20%,50%and 90%thickness reduction through cold rolling at liquid nitrogen temperature,followed by annealing at 750,950 and 1050℃for 15 min.Recrystallization occurred after annealing at 750℃,and with the increasing of annealing temperature to 950℃and 1050℃,secondary recrystallization(abnormal grain growth)and grain growth were observed.The results showed that,after 20%thickness reduction,corrosion resistance increased significantly(21.1 kΩ·cm^(2))compared with the as-received condition(3.9 kΩ·cm^(2))due to the enhancement ofγ-fiber and the creation ofΣ3 boundaries.In contrast,the corrosion resistance decreased with the increasing of thickness reduction to 90%during rolling,but still depicted higher corrosion resistance compared with the as-received specimen.After annealing the 90%cold rolled(CR)specimens at 750 and 950℃,the corrosion resistance increased in comparison with the as-received sample as a result of the more uniform microstructure,appearance of Goss and brass texture components,and grain refinement.However,significant grain growth((112±76)μm)followed by a non-uniform structure was observed after annealing at 1050℃and resulted in the lowest corrosion resistance(1.3 kΩ·cm^(2)).

Raman spectroscopy investigation of structural and textural change in C/C composites during braking

The microstructure and texture of C/C composites with a resin-derived carbon, a rough laminar （RL） pyrocarbon and a smooth laminar pyrocarbon, before and after braking tests, were investigated by Raman spectroscopy. The full width at half maximum （FWHM） of the D-band indicates the amount of defects in the in-plane lattice, while the G-to-D band intensity （peak area） ratios （lC/ID） is used to evaluate the degree of graphitization. The results show that the FWHM of D-band of sample with RL pyrocarbon changes greatly from 36 cm-1 to 168 cm 1 after braking tests, which indicates that a large number of lattice defects are produced on its wear surface. However, the graphitization degree of resin-derived carbon sample rises significantly, because the IC/1D increases from 0.427 to 0.928. Braking tests under normal loading conditions, involving high temperature and high pressure, produce a lot of lattice defects on the wear surface, and induce the graphitization of the surface. Sample with RL pyrocarbon having a low hardness is easy to deform, and has the most lattice defects on the wear surface after braking. While raw materials with resin-derived carbon have the lowest graphitization degree which rises greatly during braking.

Study on engineering geological stability of rock mass at Shanmen silver deposit

The natural balance conditions will be disturbed and produce a series of problems when mineral deposit has mined.This paper has researched the engineering rock masses have been researched in this study,structural planes,the distribution characteristics of tectonic geological factors and the stability of engineering structures according to the theory and research methods of rock mechanics,it will provide the engineering geological evidence for mining area exploited,meanwhile pledge the safety production.Shanmen silver deposit is a large epithermal deposit,it is controlled by NE to NNE strike faults.The stability of rock mass is acted on the tectonic movement and hot metalliferous brine in long-term.Especially,strength of rock mass becomes softened,muddy and loosed under the action of water,so the lower stability of rock mass is,the easier it can take place for harm of disaster threatening production safe of mining.For this reason,it is very important that drawing up a plan to lower harm for mine and protect.

Method of ameliorating the lubrication and friction performance of an engine based on different microtextures

A design of different microtextures on the surface of the crankpin bearing(CB)is proposed to ameliorate the lubrication and friction performance(LFP)of engines.On the basis of the CB s hydrodynamic lubrication model,the bearing surface of CB using different microtextures,such as wedge-shaped textures(WSTs),square textures(STs),circular textures(CTs),and combined square-circular textures(CSCTs),is simulated and assessed under various external loads of the CB at an engine speed of 2000 r/min.The pressure of the oil film,the frictional force,the force of the solid asperity contact,and the friction coefficient of the CB are used as objective functions.Results indicate that the bearing surface designed by the STs remarkably improves the CB s LFP in comparison with other structures of WSTs,CTs,and CSCTs.Particularly,the average values of the frictional force,solid asperity contact,and friction coefficient of the CB using the STs are greatly reduced by 28.5%,14.5%,and 33.2%and by 34.4%,26.3%,and 43.6%in comparison with the optimized CB dimensions and CTs,respectively.Therefore,the application of the STs on the CB surfaces can enhance the LFP of engines.

局部梯度增强的X射线图像去噪TV模型

为解决自动X射线检测设备采集的球形栅阵列(ball grid array,BGA)电路板图像常带有大量噪声的问题,根据BGA电路板焊点检测的需求和X射线图像噪声的特点,将去噪问题转化为图像的纹理滤波,在全变分(total variation,TV)优化模型的基础上,提出了结合局部梯度信息的图像去噪TV模型。首先利用图像中结构和纹理区域内的局部梯度信息差异,设计局部梯度增强的TV(local gradient enhancement TV,LGETV)模型;然后通过迭代重加权最小二乘法运算得到去噪结果图像。实验结果表明算法在BGA电路板X射线图像的去噪性能上优于多种相关算法。

Effects of strain rate and temperature on microstructure and texture for AZ31 during uniaxial compression

In order to investigate the effects of strain rate and temperature on the microstructure and texture evolution during warm deformation of wrought Mg alloy,AZ31 extruded rods were cut into cylinder samples with the dimension of d8 mm×12 mm.The samples were compressed using a Gleeble 1500D thermo-mechanical simulation machine at various strain rates(0.001,0.01,0.1,1 and 5 s- 1)and various temperatures(300,350,400 and 450℃).The microstructure and texture of the compressed samples at the same strain under different deformation conditions were studied and compared by electron backscatter diffraction(EBSD)in scanning electron microscope(SEM).The results show that the size of recrystallized grains in the deformed samples generally increases with the decrease of strain rate and the increase of temperature.After 50%reduction,most basal planes are aligned perpendicular to the compression direction at relatively high strain rate(>0.01 s- 1)or low temperature(<350℃).The optimized strain rate is 0.1 s- 1for uniaxial compression at 300℃,which produces about 80%of small grains(<5μm).

Research on the surface passivation of nanostructure-textured crystalline silicon solar cell

Nanostructure-textured solar cell owns unique properties but has some shortages especially in its fabrication and passivation.In this paper,nanostructures for crystalline silicon solar cell have been synthesized by controllable method based on silver catalyzed chemical etching.In this way,only the front surface of cell is etched and rear surface is protected.It was found that cells textured via the new method obtained equally excellent optical while superior electrical properties compared with those textured via traditional HF/AgNO3 etching.The V OC and I SC of the cell were improved by 6% and 11%,respectively.Then the cells were passivated via a bi-layer passivation(SiO2 & SiN x),in contrast to traditional SiN x passivation.It was also found that cells with new passivation exhibited improved V OC and I SC by 4% and 25%,respectively.The encouraging results can provide fundamental data for developing the nanostructure-textured crystalline silicon solar cell in following researches.

Microstructure and texture evolution of cold drawing <110> single crystal copper

By means of electron backscattering diffraction and transmission electron microscopy the microstructure and texture of drawn single crystal copper with initial orientation （110） parallel to axial direction have been investigated in the present work. In or- der to analyze the effect of initial orientation on microstructure and texture of drawn copper, the results of the drawn （110） sin- gle crystal copper wires have been compared with （100） and （111） single crystal copper wires. It is found that the grain subdi- vision of （110） single crystal is more evident than that of （100） and （111）, and the textures consisting of （111） and （100） abruptly form in the drawn （110） single crystal. At high strains, due to shear strain, the distribution of fiber textures is imho- mogenous along the radial direction of drawn （110） single crystal copper wires. （100） is near the surface and （111） is at the center. The microstructure results of drawn （110） single crystal show that at low strains, it can be characterized as two kinds of geometrically necessary boundaries with noncrystalline character. At medium strains, S bands can be observed. At high strains, lamellar boundaries form. Mean misofientation and average spacing of dislocation boundary are larger in drawn （110） single crystal, as compared with （111） and （100）. In drawn （110） single crystal with high strains, the bimodal distribution forms at lower strains than in drawn （100） single crystal, which is because the dislocation boundaries with high angle are contributed by not only the boundary between （111） and （100） fiber textures but also the boundary in （111） or （100） texture.