一种改进的势函数聚类多阈值图像分割算法

针对基于势函数聚类的多阈值图像分割算法的不足,定义了伪势的概念,并在原算法基础上提出了一种改进的图像分割算法。由伪势概念确定了伪势合并的判别方法,按照此方法,当相邻的两个峰之间的距离小于所定义的自适应模糊伪势因子时,则应该进行伪势合并。改进后的算法在计算剩余势函数时判断是否存在伪势,然后在势划分函数组的确定过程中相应地进行伪势合并计算。利用多幅图像进行了多阈值分割的仿真试验,结果表明,改进的基于势函数的多阈值图像分割算法具有更好的鲁棒性和分割效果。     

综合边缘检测和区域生长的红外图像分割方法

针对红外图像的特点,提出了一种综合应用边缘检测和区域生长方法的图像分割方法。其思路为:先对图像进行边缘提取,得到边缘像素点集;然后利用该点集的平均灰度和目标区域的连通性作为生长判决条件,采用区域生长法实现图像分割。仿真结果表明,该方法能快速准确有效地实现红外图像分割,避免了单独使用边缘提取或区域生长法进行图像分割时的典型分割错误。     

Robust multi-planar region detection for image mosaic based on hierarchical clustering mechanism

Abstract

Planar structures exist widely in the images of various scenes, and the detection of planar regions is important in many applications related to computer vision, such as image mosaic and three-dimensional reconstruction. In this paper, a robust detection method for multi-planar regions is proposed. After the feature point pairs are extracted, their preference vectors are generated in similar conceptual space. By introducing the shared nearest neighbour in clustering procedure, the feature point pairs with smaller Jaccard distance and more shared nearest neighbours simultaneously are clustered into the same planar region. Because the relationship between the feature point pairs is considered, the accuracy of the inlier probability is high. Our method can detect multi-planar regions correctly without pre-determining the number of regions, and the corresponding clustered feature point pairs can be easily utilised for image mosaic. The experimental results show the effectiveness of the proposed method.     

基于语义分割的实时车道线检测方法

车道线识别是自动驾驶环境感知的一项重要任务。近年来,基于卷积神经网络的深度学习方法在目标检测和场景分割中取得了很好的效果。本文借鉴语义分割的思想,设计了一个基于编码解码结构的轻量级车道线分割网络。针对卷积神经网络计算量大的问题,引入深度可分离卷积来替代普通卷积以减少卷积运算量。此外,提出了一种更高效的卷积结构LaneConv和LaneDeconv来进一步提高计算效率。为了获取更好的车道线特征表示能力,在编码阶段本文引入了一种将空间注意力和通道注意力串联的双注意力机制模块(CBAM)来提高车道线分割精度。在Tusimple车道线数据集上进行了大量实验,结果表明,本文方法能够显著提升车道线的分割速度,且在各种条件下都具有良好的分割效果和鲁棒性。与现有的车道线分割模型相比,本文方法在分割精度方面相似甚至更优,而在速度方面则有明显提升。

     

融合注意力机制和语义关联性的多标签图像分类

卷积神经网络在单标签图像分类中表现出了良好的性能,但是,如何将其更好地应用到多标签图像分类仍然是一项重要的挑战。本文提出一种基于卷积神经网络并融合注意力机制和语义关联性的多标签图像分类方法。首先,利用卷积神经网络来提取特征;其次,利用注意力机制将数据集中的每个标签类别和输出特征图中的每个通道进行对应;最后,利用监督学习的方式学习通道之间的关联性,也就是学习标签之间的关联性。实验结果表明,本文方法可以有效地学习标签之间语义关联性,并提升多标签图像分类效果。     

一种基于DA-GMRF的无监督图像分割方法

提出一种基于间断自适应高斯马尔可夫随机场(DA-GMRF)模型的无监督图像分割方法.针对MRF模型中的过平滑问题,利用边缘信息构造能量函数,定义了一种DA-GMRF模型.利用灰度直方图势函数自动确定分类数及分割阈值,进行多阈值分割,得到DA-GMRF模型中标记场的初始化,用Metroplis采样器算法进行标记场的优化,实现了图像的无监督分割.实验结果表明了该方法的有效性.     

基于语义分割的火车车厢位置检测研究

为了实现在煤炭定量装车站装车过程中实时检测火车车厢位置,为溜槽升降提供触发信号,设计了一种基于语义分割的火车车厢位置检测模型。以FPN （feature pyramid networks,特征金字塔网络）和ResNet101 （residual network,残差网络）为主干网络,提取并融合分辨率、语义强度不同的特征图;结合基于期望最大化（expectation maximization, EM）算法的注意力机制,构建车厢上边框语义分割模型,用于过滤特征图中的噪声,提高图像边界的语义分割精度;设计位置检测模块,计算语义分割后图像中各类别的面积及其比例和车厢上边框外接矩形高度,以获取火车车厢位置信息。结果表明,所构建的车厢上边框语义分割模型在测试集上的mIoU （mean intersection over union,均交并比）为81.21%,mPA （mean pixel accuracy,平均像素精度）为88.64%,相比未引入注意力机制的语义分割模型分别提升了3.91%和7.44%。在煤炭定量装车站现场进行的火车车厢位置检测试验结果表明,基于语义分割的火车车厢位置检测模型的检测精度满足煤炭装车过程中车厢位置检测任务的要求,这为实现煤炭定量装车系统的智能化提供了新思路。     

Improved fuzzy entropy clustering algorithm for MRI brain image segmentation

Magnetic resonance imaging (MRI) brain image segmentation is essential at preliminary stage in the neuroscience research and computer‐aided diagnosis. However, presence of noise and intensity inhomogeneity in MRI brain images leads to improper segmentation. The fuzzy entropy clustering (FEC) is often used to deal with noisy data. One major disadvantage of the FEC algorithm is that it does not consider the local spatial information. In this article, we have proposed an improved fuzzy entropy clustering (IFEC) algorithm by introducing a new fuzzy factor, which incorporates both local spatial and gray‐level information. The IFEC algorithm is insensitive to noise, preserves the image detail during clustering, and is free of parameter selection. The efficacy of IFEC algorithm is demonstrated by comparing it quantitatively with the state‐of‐the‐art segmentation approaches in terms of similarity index on publically available real and simulated MRI brain images.     

Improved quadtree image segmentation approach to region information

Images are full of information and most often, little information is desired for subsequent processing. Hence, region of interest has key importance in image processing. Quadtree image segmentation has been widely used in many image processing applications to locate the region of interest for further processing. There are also variable block-size image coding techniques to effectively reduce the number of transmitted parts. This paper presents quadtree partition technique as a pre-processing step in image processing to determine what part should be more heterogeneous than the others. It also introduces an idea to solve the problem of squared images. Finally, proposed approach is implemented and analysed. The simulation of the Matlab code of the quadtree is represented by all algorithms and the figures. Thus, achieved results are promising in the state of the art.     

基于SLIC与分水岭算法的彩色图像分割

为了克服传统分水岭算法引起的过分割问题,提出了一种基于简单线性迭代聚类(SLIC)与分水岭算法相结合的彩色图像分割算法,以获得更理想的分割效果。该算法首先利用图像复杂度计算预分割的超像素个数,并利用SLIC对原始图像进行超像素分割预处理,以减少后续处理中的冗余信息;然后,提出了一种自适应计算阈值的方法对预处理图像的梯度图像进行阈值处理,以有效去除噪声,获得较完整的轮廓信息;最后,利用分水岭分割算法对进行极小值标记提取后的图像进行分割。通过对大量图片进行实验表明,本文算法可以有效地抑制传统分水岭算法所产生的过分割问题,在LCE和GCE的对比上优于传统算法,分割质量有所提高。     

基于视觉单词和语义映射的色情图像检测算法

针对传统类型的色情图像检测方法误检率高的问题,提出了一种基于多层视觉单词的检测方法.该方法首先对色情场景的各种视觉元素建立视觉单词,然后通过这些视觉单词建立更高层的编码,包括视觉词组和兴趣区域类别,从而实现对图像不同形态级别的描述与分析.图像的识别特征由相应的编码直方图组成,并将特征映射到一个低维空间中,使图像间的语义距离与空间距离相协调.该方法在各种图像测试中都表现出出色的性能,例如在人物类图像测试中,误检率比传统方法降低了40%.实验结果证明,多层单词体系能够更高效地分析色情图像等复杂场景.     

Efficient image segmentation based on deep learning for mineral image classification

Mineral image segmentation plays a vital role in the realization of machine vision based intelligent ore sorting equipment. However, the existing image segmentation methods still cannot effectively solve the problem of adhesion and overlap between mineral particles, and the segmentation performance of small and irregular particles still needs to be improved. To overcome these bottlenecks, we propose a deep learning based image segmentation method to segment the key areas in mineral images using morphological transformation to process mineral image masks. This investigation explores four aspects of the deep learning-based mineral image segmentation model, including backbone selection, module configuration, loss function construction, and its application in mineral image classification. Specifically, referring to the designs of U-Net, FCN, Seg Net, PSP Net, and DeepLab Net, this experiment uses different backbones as Encoder to building ten mineral image segmentation models with different layers, structures, and sampling methods. Simultaneously, we propose a new loss function suitable for mineral image segmentation and compare CNNs-based segmentation models' training performance under different loss functions. The experiment results show that the proposed mineral image segmentation has excellent segmentation performance, effectively solves adhesion and overlap between adjacent particles without affecting the classification accuracy. By using the Mobile Net as backbone, the PSP Net and DeepLab can achieve a high segmentation performance in mineral image segmentation tasks, and the 15 × 15 is the most suitable size for erosion element structure to process the mask images of the segmentation models.     

Cervical image classification based on image segmentation preprocessing and a CapsNet network model

Cervical cancer is one of the most common gynecological malignancies, and when detected and treated at an early stage, the cure rate is almost 100%. Colposcopy can be used to diagnose cervical lesions by direct observation of the surface of the cervix using microscopic biopsy and pathological examination, which can improve the diagnosis rate and ensure that patients receive fast and effective treatment. Digital colposcopy and automatic image analysis can reduce the work burden on doctors, improve work efficiency, and help healthcare institutions to make better treatment decisions in underdeveloped areas. The present study used a deep-learning model to classify the images of cervical lesions. Clinicians could determine patient treatment based on the type of cervix, which greatly improved the diagnostic efficiency and accuracy. The present study was divided into two parts. First, convolutional neural networks were used to segment the lesions in the cervical images; and second, a neural network model similar to CapsNet was used to identify and classify the cervical images. Finally, the training set accuracy of our model was 99%, the test set accuracy was 80.1%, it obtained better results than other classification methods, and it realized rapid classification and prediction of mass image data.     

灰度值频数和遍历八方向的指纹图像分割算法

提出基于灰度值频数和遍历八方向的指纹图像分割算法。对于脊谷线灰度值相差较大的,利用灰度直方图上出现频数较多的两个灰度级的差值大小,判断是否为指纹前景区;脊谷线灰度值相差不大的,利用纹线的方向性,通过八个方向的模板计算在各个方向上灰度差值的大小,确定是否为指纹前景区。该方法的阈值可以根据图像自然决定,避免了人为选择阈值的困难和不准确性。对脊谷线灰度值相差较大、不大、较小的指纹图均能容易而准确地分割,只求出指纹的边界而不改变指纹图。实验表明,与常见的分割方法—方向图、方差法相比,该方法的平均误分概率大大减小,比方差法降低5.7875%,比方向图法降低5.6625%,且对指纹图像脊谷线的对比度和方向性要求不高,鲁棒性更强。     

Microscopic image segmentation based on pixel classification and dimensionality reduction

Pathological image analysis plays a significant role in effective disease diagnostics. In this article, a tool for diagnosis assistance by automatic segmentation of bone marrow images is introduced. The aim of our segmentation is to demarcate cell's component: nucleus, cytoplasm, red cells, and background. Different color spaces were used to extract color's features to profit of their complementarity. We introduce several dimensionality reduction techniques. These techniques are exemplified on a support vector machine pixel‐based bone marrow image segmentation problem in which it is shown that it may give significant improvement in segmentation accuracy and time consuming. © 2013 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 23, 22–28, 2013     

红外序列图像的支持向量机分割方法

红外序列图像的准确分割是自动目标识别的关键,而当图像背景复杂时,传统的图像分割技术往往难以满足要求,为此,提出了基于支持向量机的红外序列图像分割方法。序列图像中的部分帧被作为训练样本,通过选择适合的模型参数,运用支持向量机方法建立学习机器,将后续图像帧中的目标从复杂的背景中识别出来,从而实现红外图像分割。实际红外序列图像分割表明,基于支持向量机的图像分割方法不需要复杂的预处理和后处理工作,分割效果理想,对于小目标的图像,识别正确率可达 99%。     

基于残差和注意力网络的声呐图像去噪方法

前视声呐作为一种水下主动声呐设备常用于采集水下图像数据,然而会受到水下噪声的影响导致图像质量下降。针对这一问题,本文提出了一种基于密集残差和双通道注意力机制网络的前视声呐图像去噪方法。首先采用双通道注意力机制对声呐图像的通道信息进行提取,统计声呐图像的全局信息,输出声呐图像的噪声图;密集残差块根据噪声图和声呐图像,充分学习不同尺度上的特征信息,经过多次学习和信息传递后输出干净声呐图像。针对前视声呐图像及其噪声特点,模拟了前视声呐图像并添加瑞利分布的乘性噪声和高斯分布的加性噪声,生成模拟数据集用于网络训练和性能评估。在模拟数据集和真实数据集的实验中表明,本文方法能够有效去除噪声,保留图像细节。

     

一种基于改进PCNN模型的图像分割方法

通过对传统脉冲耦合神经网络(PCNN)模型的改进,在模型的输入端加入目标区域的边缘数据,使最高灰度级不同的非连通神经元同期点火,实现了多目标区域同时分割。给出了影响同期点火激励范围的主要参数β的自动设定方法,并设计了基于图像最大熵准则的自动分割算法。用分割精度评价准则验证了所提出方法的有效性。实验证明,对于低噪声污染的图像,改进的PCNN模型在多目标识别中的正确接受率达到95%以上,明显优于经典的Fastlinking模型。     

基于多阈值单水平集方法的医学图像分割

针对Chan-Vese的无边界主动轮廓模型(CV模型)只能区分前景与背景的缺点,提出了一种基于多阈值单水平集的医学图像分割方法,并将此方法应用于微创手术的预处理中.由于医学图像结构复杂,具有器官轮廓多连接等特点,因此使用常规的水平集方法进行分割往往不能取得理想的效果,而该方法采用修改目标泛函的方式引入多类分割,具有多区域分割的特点,只需经过一次单水平集的迭代循环,即可将图像根据灰度不同划分为多个区域,具有精确、快速等优点.对不同的合成图像和医学图像的实验结果表明,该方法实现了快速精确的多区域分割,能很好地提取到医学图像中的骨骼轮廓,分割效果达到了预期水平.     

基于改进C-V模型的超声图像分割方法

图像分割是超声医学图像学中的难题之一。改进的Chan-Vese（C-V）法加入了约束符号距离函数的能量项，避免了演化时候的重新初始化。在改进C-V模型的基础上，首先借用分水岭中的思想，找到分割目标的近似轮廓，并以此轮廓生成符号距离函数，然后采用改进的C-V法进行超声图像分割。实验表明，改进的方法有更高的精准度和对多目标分割的能力。