New algorithms for preserving edges in low‐bit‐rate wavelet‐based image compression

With the growing importance of low‐bandwidth applications such as wireless access to the Internet, images are often sent or received at low bit rates. At these bit rates, images suffer from significant distortion and artifacts, making it difficult for those viewing the images to understand them. In this paper, we present two progressive compression algorithms that focus on preserving the clarity of important image features, such as edges, at compression ratios of 100:1 and higher. The first algorithm transmits a standard SPIHT bit stream and then detects the location of edges in the compressed image. The decoder applies a linear edge‐enhancement procedure to improve the clarity of the encoded edges. The second algorithm extracts the locations of straight‐line edges in the image at the encoder, and the decoder applies edge extraction, combination, and a linear edge‐enhancement procedure to improve the clarity of the edges. With both algorithms, features in the images that may be important for recognition are well preserved, even at very low bit rates. © 2012 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

多尺度压缩求导的一维图像边缘检测

为了更加精确地计算塑料薄膜缺陷的宽度,在基于小波算法的多尺度变换基础上,提出了一种用于一维图像不同尺度间的压缩求导边缘检测方法.小波变换提供图像的多尺度描述,将压缩求导应用于不同尺度间的小波域,从而可将各尺度的图像信息更加有效地合成,得到最优的边缘检测效果.该方法比以往的单纯直线拟合方法具有更好的精确度,可以有效地增强边缘和抑制噪声.对一维图像进行的实验证明该方法是正确和有效的.     

基于小波变换的新型SURF图像拼接方法

针对传统图像拼接算法特征点计算量大、耗时较长等问题,提出了一种基于小波变换的新型加速鲁棒特征算法(SURF)图像拼接方法。首先通过Haar小波函数对图像进行二阶分解以获取图像低频成分,并利用小波梯度矢量对低频图像重合区域进行特征点提取,从而实现低频图像下快速获得特征点的变换参数以指导高频图像下的特征点提取;在此基础上,提出一种SURF图像匹配改进算法,利用特征点约束的单向匹配和方向一致等性质,有效剔除误匹配点对,以提高特征点匹配精度和实时性。最后,通过两组实验验证了所提出方法的有效性和可行性。     

基于多尺度Gabor滤波器的角点检测

为了克服在不同图像上的尺度选择问题,提出了一种基于边缘轮廓线的多尺度Gabor滤波器的角点检测算法。该算法首先利用Canny边缘检测算子提取图像的边缘轮廓;进而用一组构建好的4个尺度8个方向Gabor滤波器的虚部对图像进行平滑,并计算每个像素在其相同尺度下各个方向上Gabor滤波器虚部响应的归一化的和;最后将每个边缘像素点在所有尺度下的乘积作为新的角点测度, 当角点测度大于预设阈值时,则认定该点为角点。将实验结果与经典的Harris、CPDA和He&Yung角点检测算法进行比较,提出的算法在检测准确率、定位误差、噪声稳健性性能指标上,都取得了更好的结果。     

Automated registration of brain images using edge and surfacefeatures

We present an automated multimodality (CT & MRI) registration algorithm based on hierarchical feature extraction. Two kinds of shape representations-edge and surface-are extracted hierarchically from different image modalities. The registration then is performed using least-squares matching of the user-specified (but automatically extracted) corresponding features. In our implementation, the 3-D version of the Canny edge detector is employed in the extraction of corresponding edge information. An automatic segmentation algorithm is introduced to extract the corresponding surfaces from the edges efficiently. The geometric matching of those extracted shape features then is performed using the iterative closest-point matching method     

输电线路覆冰图像边缘提取方法研究

A new method of measuring the icing thickness of transmission lines on-line is proposed in this paper.In this method,the pictures of transmission lines which are photoed by the camera on the iron tower are processed immediately to extract the edges of the transmission line conductor and transmission line insulators.The icing thickness can be gained by comparing the edges of the iced transmission line and the uniced one.Two icing image edge extraction methods are described in detail,that is,a method based on the combination of the wavelet transform and the floating threshold method and a method based on the combination of the optimal threshold method and the mathematical morphology transform.The icing images from the artificial climatic chamber and transmission lines are used to test the methods above.The results show that the method based on the wavelet transform and the floating threshold method does well in the extraction of relatively smooth edges,such as glaze icing on conductor and icing on the insulator;meanwhile,the method based on the optimal threshold method and the mathematical morphology transform does well in the edge extraction of icing on the conductor,especially the opaque rime icing on the conductor with complicated edges.     

基于颜色空间变量的输电线图像分类及特征提取

针对输电线图像背景复杂多变和单一图像处理方法难以有效处理各种背景类型输电线图像的问题,提出一种基于颜色空间变量的输电线图像分类及特征提取方法。首先根据输电线图像各颜色空间的变量值与图像特征之间的关系对图像进行分类。然后根据不同类别图像特征选用适合的滤波方法通过两次滤波结合去噪,并采用自适应直方图分段均衡化增强图像对比度。通过对Otsu算法得出的阈值进行线性变换确定canny边缘检测参数,提取输电线路边缘。最后根据输电线形状特征和概率霍夫直线变换与形态学运算提出一种边缘优化方法,较好地去除非输电线边缘。结果表明:该方法可以有效处理各种背景类型图像,为输电线路图像智能化处理提供了一种新的思路和方法。     

基于小波变换和边缘模板相结合的序列B超图像边缘提取

由于B超图像的分辨率差、灰度级别少等特点,用传统的边缘检测难以得到满意的效果。本文采用了基于离散二进小波变换的门限选取法,通过检测图像直方图不上波变换的零交叉点来得到图像的边缘;进而又构造出一种边缘模板,利用小波变换和边缘模板相结合的方法提取出序列B超图像的边缘。     

Comparative performance of principal component analysis, gabor wavelets and discrete wavelet transforms for face recognition

This paper compares the performance of face recognition systems based on principal component analysis (PCA), Gabor wavelets (GW) and discrete wavelet transform (DWT). The three techniques are implemented in the MATLAB programming environment, and their performance is investigated using frontal facial images from the FERET database. The images are preprocessed to yield a standardized image used for identification. PCA produces an orthonormal basis for the image space that extracts the dominant facial features, providing exceptional recognition performance. The GW technique is modelled after biological experiments and is used to filter spatial-frequency features of the image at key points of the face. The DWT is investigated for its potential use in facial-feature extraction and is also applied to rotated versions of the facial image, thereby increasing the directional filtering capability. A face similarity measure that uses the extracted features provides recognition that is robust against variations in illumination.     

基于深度学习与不平衡样本集的输电线路故障分类

针对输电线路各类型故障样本间的数量不平衡会造成人工智能算法对故障中的少数类样本识别精度不足的问题,提出了一种基于Borderline-SMOTE（BSMOTE）算法与卷积神经网络（CNN）相结合的输电线路故障分类方法。该方法首先利用BSMOTE算法对位于分类边界上的少数类样本进行过采样合成处理,改善样本间的不平衡度,然后将所提取的一维故障电流信号样本重构成二维灰度图像数据形式,并在Pytorch深度学习框架下搭建了CNN网络模型,利用模型的自主学习能力对灰度图像进行特征自提取与辨识,减少传统人工设计特征提取的工序,完成对输电线路故障类型的分类。实验结果表明该模型能够提高对少数类故障样本的识别能力,准确地判断故障类型,并对噪音具有较强的抗干扰能力。     

基于小波分析的图像检索技术

随着大量数字图像数据库的出现,基于内容的图像检索技术成为研究热点。本文针对小波的多尺度分析特性,描述了一种基于小波分析的图像检索算法,在小波变换域内提取出图像的显著特征点,然后提取显著特征点的颜色和纹理特征,颜色选取HSV空间,纹理特征用Gabor小波幅值的均值与方差表示,通过相似性匹配实现图像的检索。最后,本文通过实验证实了这种方法具有较高的检索精度,与基于角点检测方法的图像检索相比,检索效果与人的视觉更加相符。     

基于卷积神经网络的非侵入式负荷监测方法

从深度学习与边缘计算的角度,对适用于电力物联网的非侵入式负荷监测方法展开了研究.针对NILM系统在物联网场景下的部署问题,提出了一种新的边缘计算架构,并讨论了各组成部分的任务分配.针对负荷激活在线提取问题,提出了基于离散度和用电行为规律分析的激活判断策略;针对低频采样下的负荷特征问题,提出了一种可自动提取激活特征并识别...     

基于无人机图像处理的输电线路提取技术研究

为进一步减轻输电线路进行定期检查、巡视的任务,文章提出了一种利用智能化无人机巡检技术,对航拍图像进行线路的提取和跟踪。采用直方图均衡化及图像滤波对航拍图像进行预处理,解决了航拍图像光照强度以及背景对输电线路元素提取的干扰;采用LSD算法实现了线路边缘的提取,在去除图像背景信号的基础上使用Hough变换数学算法实现了输电线路的准确连接;分别采用粒子滤波和扩展卡尔曼滤波两种图像跟踪方法对航拍视频进行线路跟踪,通过建立输电线运动模型,用仿真软件对其进行识别,两种方法的检测准确度分别为95.34%和94.72%,证实文章处理算法可实现输电线路的提取和跟踪。     

基于改进Faster-RCNN的输电线巡检图像多目标检测及定位

针对输电线巡检图像受光线、环境和拍摄角度等因素影响,图像中的电气设备呈现低分辨率和多形态化特征的问题,提出一种基于改进Faster-RCNN的巡检图像多目标检测及定位方法。该方法首先通过区域建议策略网络生成若干目标候选区域;然后基于实际巡检图像样本库,对卷积神经网络进行训练,以改善参数学习效果;最后利用正则化方法优化参数权重,提高检测速度,得到适应巡检图像多形态化特征的改进型Faster-RCNN模型。实际场景数据集测试结果表明,相比于数字图像处理、浅层机器学习、单阶法、双阶法、Mask-RCNN和Local Loss目标检测方法,所提改进型Faster-RCNN能够在不同分辨率和不同位置角度的巡检图像场景下保持较高的识别精度和速度,具有较高的工程实用价值。     

Robot vision with cellular neural networks: a practical implementation of new algorithms

Cellular neural networks (CNNs) are well suited for image processing due to the possibility of a parallel computation. In this paper, we present two algorithms for tracking and obstacle avoidance using CNNs. Furthermore, we show the implementation of an autonomous robot guided using only real‐time visual feedback; the image processing is performed entirely by a CNN system embedded in a digital signal processor (DSP). We successfully tested the two algorithms on this robot. Copyright © 2006 John Wiley & Sons, Ltd.     

基于提升小波变换的指针式仪表图像边缘检测

为解决工业中对指针式仪表的自动检定,针对自动检定系统中所涉及的图像边缘检测技术难点,分析了现有传统算法的不足之处,提出一种有效的基于提升小波变换的边缘检测算法。该算法利用提升小波变换尺度的低通作用,避免受高频噪声影响,在传统边缘检测算法的基础上,提取低频轮廓。利用小波系数的方向性,结合方向性边缘检测算子,获得高频边缘信息,最后利用小波重构获得准确清晰的图像边缘,为后续Hough变换准确的提取指针提供了有利保障。实验结果表明,该算法得到的图像边缘较传统Canny算法更清晰准确,无虚假边缘,应用于仪表自动检测系统中是可行的。     

基于神经网络的CT脑血管图像边缘检测算法

CT脑血管医学图像的三维重构都是源自二维断层扫描,脑血管边缘特征向量的提取是图像处理的关键步骤。为提高边缘特征的提取和保证三维重建图像的质量,在分析了某些常用的边缘检测算法性能基础上,同时结合CT脑血管图像的像素结构特点,将SA_SOFM神经网络算法成功地用于对CT脑血管图像的边缘特征信息提取中。并对算法进行有效的改进,基于真实图像的实验表明该算法提高了边缘特征信息的精度和鲁棒性。     

基于Iradon-CNN的变压器局部放电状态识别方法

为解决变压器局部放电故障所带来的安全隐患,提出了一种基于逆拉冬变换(Inverse Radon transform,Iradon)-卷积神经网络(Convolutional Neural Networks,CNN)的变压器局部放电信号图像识别方法。针对三种故障进行了局部放电实验,首先通过共振稀疏分解对局部放电信号进行分解,获取低共振分量,然后将其转换成Iradon图像,最后利用CNN自适应地提取Iradon图像的特征信息。结果表明,该方法能够准确提取信号特征,具有强大的数据处理和识别功能,并为变压器局部放电状态的识别提供了丰富的信息,提高了学习效果和识别精度。     

Compact CMOS implementation of a low‐power,current‐mode programmable cellular neural network

We report on the design and characterization of a full‐analog programmable current‐mode cellular neural network (CNN) in CMOS technology. In the proposed CNN, a novel cell‐core topology, which allows for an easy programming of both feedback and control templates over a wide range of values, including all those required for many signal processing tasks, is employed. The CMOS implementation of this network features both low‐power consumption and small‐area occupation, making it suitable for the realization of large cell‐grid sizes. Device level and Monte Carlo simulations of the network proved that the proposed CNN can be successfully adopted for several applications in both grey‐scale and binary image processing tasks. Results from the characterization of a preliminary CNN test‐chip (8×1 array), intended as a simple demonstrator of the proposed circuit technique, are also reported and discussed. Copyright © 2001 John Wiley & Sons, Ltd.