基于平移不变小波阈值算法的噪声图象估计

在小波基下简单的阈值算法几乎是最优的非线性估计, 若信号是分片光滑函数, 小波阈值方法所得的风险就接近于最佳值. 采用标准的小波阈值估计可较大幅度的提高PSNR, 但由于图像的边缘被Gibbs类的振荡所影响, 图像视觉质量较差. 本文采用改进的平移不变阈值算法可获得更高的PSNR, 更重要的是该方法明显地改善了图像的视觉质量.     

基于回波平移序列的快速三维磁共振温度成像方法

高强度聚焦超声（HIFU）是一种无创的热消融疗法，为保证其安全性和有效性，需要一种精度高、速度快的测温方法在其治疗过程中对温度进行监控.基于质子共振频率位移（PRFS）的磁共振温度成像（MRT）对温度具有较高的灵敏度，且与温度具有良好的线性关系，因此常被用于引导HIFU治疗.然而在实际应用中，HIFU治疗的最大隐患在于可能造成表皮灼伤，并且灼伤区域可能与焦点区域相隔较远.因此MRT的监控范围十分重要.本文基于三维回波平移成像序列，结合可控混叠的空间并行成像技术，实现了时间分辨率为3 s的快速三维温度成像.为了验证该方法的精度，本文首先设计了仿体降温实验，利用光纤温度计验证回波平移序列测温的准确度和精确度.然后在室温条件下扫描离体猪肉组织，对比加速前后的MRT的测温精确度.在HIFU加热条件下扫描离体猪肉组织，对比加速前后的MRT的测温准确度.结果显示，本文提出的方法可以在3 s内完成三维温度精准测量，对于HIFU治疗的安全监控具有重要意义.     

基于双树复数小波变换的图像去噪方法

在双树复数小波的基础上引入平移不变的重要思想,使其具有真正意义上的平移不变性.提出了一种基于平移不变双树复数小波变换的图像去噪方法.实验结果表明该方法能有效提高图像去噪效果及图像边缘检测能力.     

基于平移不变剪切波变换域图像融合算法

针对传统基于多尺度变换的图像融合方法存在的缺点,提出了一种基于平移不变剪切波变换域的自适应图像融合新方法.首先,使用平移不变剪切波变换对源图像进行分解,得到低频子带及方向带通子带系数.然后,对于低频子带系数采用梯度域奇异值分解方法估计图像的局部结构信息,提出了基于提取的特征与S函数的可变加权融合策略;对于各方向带通子带系数,提出了一种基于改进的拉普拉斯能量和匹配的“加权平均”和选择相结合的系数选择策略.最后,对得到的融合系数进行逆变换得到融合图像.通过实验可以发现相比于传统的图像融合方法,本文方法得到了更高的客观指标,融合图像视觉效果更好.     

基于平移不变剪切波变换域图像融合算法

针对传统基于多尺度变换的图像融合方法存在的缺点,提出了一种基于平移不变剪切波变换域的自适应图像融合新方法.首先,使用平移不变剪切波变换对源图像进行分解,得到低频子带及方向带通子带系数.然后,对于低频子带系数采用梯度域奇异值分解方法估计图像的局部结构信息,提出了基于提取的特征与S函数的可变加权融合策略;对于各方向带通子带系数,提出了一种基于改进的拉普拉斯能量和匹配的"加权平均"和选择相结合的系数选择策略.最后,对得到的融合系数进行逆变换得到融合图像.通过实验可以发现相比于传统的图像融合方法,本文方法得到了更高的客观指标,融合图像视觉效果更好.     

基于小波变换的磁共振图像去噪

去噪是图像处理中的一个非常重要的问题. 传统去噪方法在降低噪声的同时会模糊图像的细节,基于小波变换的图像去噪方法能在降低图像噪声的同时较好地保持图像的细节,目前基于小波变换对图像进行去噪已经成为最有前途的去噪方法之一. 本文综合小波域去噪的文献,对小波域的磁共振图像去噪方法进行了综述,详细介绍了各种小波域的磁共振图像去噪方法,并对他们的性能进行了总结,对小波域去噪方法的发展趋势进行了展望.     

高光谱遥感图像的小波去噪方法

高光谱遥感图像是由二维空间信息和一维光谱信息组成的三维数据。普通的去噪方式通常是分别对空间信息或光谱信息进行去噪,其主要缺点是忽视了高光谱图像强烈的谱间相关性和图谱合一的特点。针对这些特点,文章提出一种基于小波变换的高光谱遥感图像去噪方法。该方法对各波段高光谱图像逐一进行二维小波变换,根据含噪声大的波段与噪声小的波段的波长关系,对小噪声波段的高频系数加权求和,代替噪声大的波段的高频系数,通过小波逆变换得到去噪后的重构图像。该方法运算速度快,能有效地降低噪声。对机载可见红外成像光谱仪数据(AVIRIS)实验表明,与经典的BayesShrink图像去噪方法相比,方法重构图像的信噪比(SNR)高出3.8～10.6 db,节省运算时间一半以上。     

电力设备红外图像的小波阈值去噪方法研究

对于电力设备的红外图像自动检测系统,图像去噪是非常关键的。针对传统的小波硬、软阈值函数在去噪时存在的不足,在传统小波硬、软阈值函数的基础上对其进行了改进。改进的阈值函数克服了硬阈值函数不连续的缺陷,改善了软阈值函数具有恒定偏差的不足,并引入了两个变量,具有一定的灵活性。同时还使用了一种新的分层阈值选择函数代替统一阈值方法,以改善实际应用的效果。实验结果表明:改进的小波阈值去噪方法在视觉效果、峰值信噪比和均方误差方面都优于传统的硬、软阈值去噪方法;改进的小波阈值去噪方法可以运用到红外图像自动检测系统中,使系统具有更好的去噪效果。     

基于自适应无下采样提升框架的Stokes图像获取方法

根据成像偏振探测中stokes参数图像获取的原理和方法，分析了传统求取stokes参数图像方法在保留图像细节信息和处理输入图像微小平移等方面的不足，在此基础上提出了基于自适应平移不变小波分解的stokes参数图像求取方法，并用无下采样的提升框架来执行平移不变小波分解以提高算法的灵活性和计算速度.针对不同的stokes参数图像求取方法，用均值、方差、熵、边界能量对融合结果进行评价和分析.通过仿真，表明本文方法提高了stokes参数图像的质量并改善了stokes参数图像计算方法对输入图像平移的敏感性.     

基于最大熵原理的小波去噪方法

小波变换去噪中最关键的问题是如何确定小波系数的阈值,使其能够将与噪声和信号相对应的小波系数合理地区分开来。根据概率论的基本原理可以推断,随机序列的细节小波变换系数符合正态分布。基于此结论,可以利用最大熵原理确定一个阈值,使得绝对值小于此阈值的小波系数组成的序列符合正态分布。该阈值在统计意义下能够最佳地区分信号与噪声的小波系数。采用光谱数据的仿真分析以及与其他方法的对比实验证明,这种最大熵小波去噪方法不仅在提高信噪比方面显示出了其优势,而且去噪效果不易受信噪比变化的影响。     

MRI dynamic range and its compatibility with signal transmission media

As the number of MRI phased array coil elements grows, interactions among cables connecting them to the system receiver become increasingly problematic. Fiber optic or wireless links would reduce electromagnetic interference, but their dynamic range (DR) is generally less than that of coaxial cables. Raw MRI signals, however, have a large DR because of the high signal amplitude near the center of k-space. Here, we study DR in MRI in order to determine the compatibility of MRI multicoil imaging with non-coaxial cable signal transmission. Since raw signal data are routinely discarded, we have developed an improved method for estimating the DR of MRI signals from conventional magnitude images. Our results indicate that the DR of typical surface coil signals at 3 T for human subjects is less than 88 dB, even for three-dimensional acquisition protocols. Cardiac and spine coil arrays had a maximum DR of less than 75 dB and head coil arrays less than 88 dB. The DR derived from magnitude images is in good agreement with that measured from raw data. The results suggest that current analog fiber optic links, with a spurious-free DR of 60–70 dB at 500 kHz bandwidth, are not by themselves adequate for transmitting MRI data from volume or array coils with DR 90 dB. However, combining analog links with signal compression might make non-coaxial cable signal transmission viable.     

基于多小波变换的图像去噪新方法

离散多小波在图像去噪和图像融合方面有突出的效果.利用GHM多小波,对图像去噪中阈值的选取给出了具体的方法,对单小波与多小波去噪方法进行了比较.结果表明,多小波在图像去噪的效果上明显优于单小波.     

The short TI inversion recovery sequence--an approach to MR imaging of the abdomen

Initial clinical experience with magnetic resonance imaging (MRI) of the abdomen using short TI inversion-recovery (STIR) pulse sequences is described and compared with X-ray CT in a variety of abdominal disease. The extent of abnormality shown with MRI was greater than that with CT in 21 of 30 cases and equal in 9 cases. Lesion contrast was greater with MRI in 15 cases, equal in 14 and less in 1. The level of artefact was equal in 27 cases and greater with MRI in 3 cases. The STIR pulse sequence has significant advantages in producing high soft-tissue contrast, controlling respiratory artefact, avoiding confusion with intra-abdominal fat and identifying bowel loops.     

基于多小波变换的红外目标探测与识别

针对光电联合变换相关器目标识别的实际应用,对待测红外目标图片进行多小波变换,并利用模极大值法提取其边缘.通过获取更多的轮廓信息,从而提高对目标的识别能力.计算机模拟了常用于红外目标处理的多小波GHM和SA4,实验结果表明:基于GHM多小波提取的边缘能获取大量的图像轮廓信息,其识别结果明显优于SA4多小波.将目标原图的光学相关探测结果与基于GHM多小波提取的边缘图像光学相关探测结果进行比较发现,经多小波预处理后的边缘图像能有效增强相关峰强度.     

Liver schwannoma: findings on MRI

Magnetic resonance imaging features of benign liver schwannoma in a 52-year-old woman are described. An oval shaped, 4.4×3.6×2.9-cm lesion was located in Segment 7 of the right hepatic lobe. The lesion was hypointense on T₁-weighted images and mixed hypo- and hyperintense on T₂-weighted images. On serial contrast-enhanced images, the lesion revealed gradually increasing centrilobular enhancement. The tumor was surgically removed thereafter.     

Denoising 3D MR images by the enhanced non-local means filter for Rician noise

The non-local means (NLM) filter removes noise by calculating the weighted average of the pixels in the global area and shows superiority over existing local filter methods that only consider local neighbor pixels. This filter has been successfully extended from 2D images to 3D images and has been applied to denoising 3D magnetic resonance (MR) images. In this article, a novel filter based on the NLM filter is proposed to improve the denoising effect. Considering the characteristics of Rician noise in the MR images, denoising by the NLM filter is first performed on the squared magnitude images. Then, unbiased correcting is carried out to eliminate the biased deviation. When performing the NLM filter, the weight is calculated based on the Gaussian-filtered image to reduce the disturbance of the noise. The performance of this filter is evaluated by carrying out a qualitative and quantitative comparison of this method with three other filters, namely, the original NLM filter, the unbiased NLM (UNLM) filter and the Rician NLM (RNLM) filter. Experimental results demonstrate that the proposed filter achieves better denoising performance over the other filters being compared.     

基于Bayesian估计和Wiener滤波的阈值去噪方法

基于阈值的小波域去噪方法的核心是阈值选取,通过最小化一个Bayesian风险函数得到一种自适应阈值,结合线性滤波中具有代表性的Wiener方法,提出了一种新的去噪方法.实验结果表明,该方法能有效地去除图像中的白噪声,同时还能较好地保留图像的边缘信息,其效果优于目前的一些小波阈值去噪方法.     

Diagnostic value of breast MRI for predicting metastatic axillary lymph nodes in breast cancer patients: diffusion-weighted MRI and conventional MRI

Purposes

To evaluate the diagnostic value of diffusion-weighted MRI (DWI) and combination of conventional MRI and DWI to predict metastatic axillary lymph nodes in breast cancer.

Materials and methods

Two hundred fifty-two breast cancer patients with 253 axillae were included. The morphological parameters on axial T2-weighted images without fat saturation and apparent diffusion coefficient (ADC) values were retrospectively analyzed. An independent t-test/chi-square test and receiver operating characteristics (ROC) curve analysis were used.

Results

On conventional MRI, short and long axis length, maximal cortical thickness, relative T2 value, loss of fatty hilum (p < 0.001 for each), and eccentric cortical thickening (p < 0.003) were statistically significantly different between the metastatic and nonmetastatic groups. The short axis to long axis ratio was not a statistically significant parameter. The ADC value was significantly different between the 2 groups, with an AUC that was higher than that of conventional MR parameters (AUC, 0.815; threshold, ≤ 0.986 × 10–3 mm²/sec; sensitivity, 75.8%; specificity, 83.9%). Using the adopted thresholds for each parameter, a total number of findings suggesting malignancy of 4 or higher was determined as the threshold, with high specificity (90.1%).

Conclusion

Using conventional MRI and DWI, we can evaluate the axilla in breast cancer with high specificity.     

Renal subcapsular pseudocyst: findings on MRI

Pancreatic pseudocysts are common complications of pancreatitis. They may occur in unusual places as a result of spread along the path of least resistance. Penetration of Gerota's fascia and spread into the renal subcapsular space are uncommon. Their appearance on magnetic resonance imaging has not been previously documented. Given the increased use of magnetic resonance cholangiopancreatography for the investigation of patients with pancreatitis, the recognition that a complicated renal cyst in this clinical context could represent a benign condition is important.     

基于神经网络和小波变换的MRI图像去噪方法

过去10年中,小波变换在图像去噪中取得了很大的成功.人们提出了多种适用于小波去噪的阈值方法,而这些方法就是希望能够正确地反映有噪声小波系数与无噪声小波系数之间的映射关系.基于这种想法,我们提出一种在小波域中利用神经网络寻找这种映射关系的图像去噪新方法.我们把该方法应用于不同噪声分布的磁共振图像的去噪,取得了良好的效果.