舌诊图像点刺和瘀点的识别与提取

计算机舌诊系统中,点刺和瘀血点是重要的舌象。基于斑点检测、支持向量机(SVM)和K-均值聚类算法,提出了对舌诊图像中点刺和瘀点的识别及提取方法。首先利用SimpleBlobDetector斑点检测算法检测斑点,并提取出斑点数量、大小和分布等特征值生成特征向量,再使用SVM进行点刺(瘀点)舌象识别。点刺(瘀点)提取同样基于斑点检测算法,提取斑点颜色特征,使用K-均值聚类将斑点聚类为多个小类簇,定义基于加权颜色空间距离的判别函数,将聚类结果同第一次斑点检测的结果对比,得到正类和负类,最终提取出点刺和瘀点。利用该方法进行实验,识别正确率达到97.4%,提取误检率为6.0%,漏检率为10.1%,表明了本方法的有效性和应用价值。     

一种用于复杂目标感知的视觉注意模型

针对自然场景图像中复杂结构目标的快速定位问题,提出一种新的视觉注意模型。对目标进行学习提取显著性图斑,将图斑的特征信息、异质图斑之间的相对位置关系引入视觉注意过程,采用基于图匹配的图斑搜索策略合并与目标特征相似的异质图斑,从而获得注意焦点。与自底向上的视觉注意模型进行实验对比,结果表明该模型能引入复杂结构目标的特征信息和结构信息,降低无效关注次数,提高视觉注意的效率。     

An adaptive water flow model for binarization of degraded document images

In this paper, we present an adaptive water flow model for the binarization of degraded document images. We regard an image surface as a three-dimensional terrain and pour water on it. The water finds the valleys and fills them. Our algorithm controls the rainfall process, pouring the water, in such a way that the water fills up to half of the valley’s depth. After stopping the rainfall, each wet region represents one character or a noisy component. To segment each character, we labeled the wet regions and regarded them as blobs; since some of the blobs are noisy components, we use a multilayer Perceptron to label each blob as either text or non-text. Since our algorithm classifies the blobs instead of pixels, it preserves stroke connectivity. After several experiments, the proposed binarization algorithm demonstrated superior performance against six well-known algorithms on three sets of degraded document images. The main superiority of our algorithm is on document images with uneven illumination.     

Boundary-aware texture region segmentation from manga

Computational Visual Media - Due to the lack of color in manga (Japanese comics), black-and-white textures are often used to enrich visual experience. With the rising need to digitize manga,...     

基于多层次空间注意力的图文评论情感分析方法

随着社交网络的不断普及,相对于传统的文字描述,人们更倾向于发布图文结合的评论来表达自己的情感与意见。针对图文情感分析方法中仅考虑图文间的高级语义联系,而较少注意图片的低层次情感特征以及中层美学特征与文本情感之间关联性的问题,提出了一种基于多层次空间注意力（MLSA）的图文评论情感分析方法。所提方法以文本内容为驱动,使用MLSA设计图像与文本之间的特征融合方法,该特征融合方法不仅关注与文本相关的图像实体特征,而且充分利用图像的中层美学特征和低层视觉特征,从而从多个不同角度挖掘图文之间的情感共现。在两个公开的图文情感数据集MVSA_Single和MVSA_Multi上,该方法的分类效果相对于对比方法中最优的方法的分类效果在准确率上分别提高了0.96和1.06个百分点,在F1值上分别提高了0.96和0.62个百分点。实验结果表明,综合分析文本特征和图像特征之间的层次化联系能有效地增强神经网络捕捉图文情感语义的能力,从而更准确地预测图文整体的情感。     

Recursive Bayesian fire recognition using greedy margin-maximizing clustering

Vision-based fire detection is a challenging research area, since the visual features of fire dynamically change due to several factors such as weather conditions. In this paper, we propose a novel fire detection approach in which detected fire-candidate blobs are categorized as fire or non-fire under recursive Bayesian estimation. By employing the recursive estimation, we attempt to deal with fire characteristics that are dynamic as well as spatiotemporally continuous in a hidden Markov process. More specifically, for each detected fire-candidate blob, future beliefs about hidden classes are predicted and corrected by the most recent beliefs and observations of the blob. This is repeated during the lifetime of the blob. In this framework, to reduce the Bayes error in classification, we devised the greedy margin-maximizing clustering algorithm. This algorithm learns color clusters to model the feature space while attempting to maximize the in-cluster margins within a class and between classes. To further improve the detection accuracy, we developed two methods, $\epsilon $ -time delayed decision and on-line learning of transition probability. These were invented to suppress false alarms caused by temporary fire-like instances and to determine the current class by considering the majority of previous classification results. Experiments and comparative analyses with two contemporary approaches are conducted for various fire situations. The results show that the proposed approach is superior to the previous approaches in detecting fire and reducing false alarms. Furthermore, the proposed approach is shown to be competitive in applications to real environments.     

A new feature-preserving nonlinear anisotropic diffusion for denoising images containing blobs and ridges

Blobs and ridges underlie many important features in biological, biometric and remote sensing images. These images are likely to be corrupted by noise, such as live cells in fluorescent biological images, ridges and valleys in fingerprints and moving targets in synthetic aperture radar and infrared images. In this paper we present a diffusion method for denoising low-signal-to-ratio images containing blob and ridge features. A commonly used denoising method makes use of edge information in an image to achieve a good balance between noise removal and feature preserving. However, if edges are partly lost to a certain extent or contaminated severely by noise, such an approach may not be able to preserve these features, leading to loss of important information. To overcome this problem, we propose a novel second-order nonlocal derivative as a robust blob and ridge detector and incorporate it into a diffusion process to form a novel feature-preserving nonlinear anisotropic diffusion model. Experiments show that the new diffusion filter outperforms many popular filters for preserving blobs and ridges, reducing noise and minimizing artifacts.     

Objects based change detection in a pair of gray-level images

The goal of the presented change detection algorithm is to extract objects that appear in only one of two input images. A typical application is surveillance, where a scene is captured at different times of the day or even on different days. In this paper we assume that there may be a significant noise or illumination differences between the input images. For example, one image may be captured in daylight while the other was captured during night with an infrared device. By using a connectivity analysis along gray-level technique, we extract significant blobs from both images. All the extracted blobs are candidates to be classified as changes or part of a change. Then, the candidate blobs from both images are matched. A blob from one image that does not satisfy the matching criteria with its corresponding blob from the other image is considered as an object of change. The algorithm was found to be reliable, fast, accurate, and robust even under extreme changes in illumination and some distortion of the images. The performance of the algorithm is demonstrated using real images. The worst-case time complexity of the algorithm is almost linear in the image size. Therefore, it is suitable for real-time applications.     

Balloon extraction from complex comic books using edge detection and histogram scoring

Extracting balloons has been a challenge in digital comics, in particular for complex comic books. In fact, current algorithms are seemingly adequate for simple comics, but not for complex ones. This paper proposes an algorithm based on the Sobel operator plus a modified flood fill operator to identify and extract balloons in comic book pages. Unlike other approaches, our algorithm applies to any sort of comics, no matter they are simple or complex comics, without making any assumptions regarding the line continuity of the image, color depth of the image, orientation of the text, and language. Experimental results show that our algorithm significantly improves the rate of correctly detected balloons, and simultaneously decreases the number of false positives, when compared to other algorithms.     

视频字幕的自动检测与去除

由于视频中固化的字幕影响了不同语种间视频的交流和处理,为此提出了一种基于CEMA算法和纹理修复技术的自动检测与去除视频内字幕的方法。首先,运用CEMA算法检测出视频中的字幕,然后,结合纹理修复技术,将检测出来的字幕从原图中去除,同时,恢复原图中被字幕所遮挡的背景区域。实验结果表明,该方法能较好地检测和去除视频图像内的字幕。     

Gaussian mixture modeling and learning of neighboring characters for multilingual text extraction in images

This paper proposes an approach based on the statistical modeling and learning of neighboring characters to extract multilingual texts in images. The case of three neighboring characters is represented as the Gaussian mixture model and discriminated from other cases by the corresponding ‘pseudo-probability’ defined under Bayes framework. Based on this modeling, text extraction is completed through labeling each connected component in the binary image as character or non-character according to its neighbors, where a mathematical morphology based method is introduced to detect and connect the separated parts of each character, and a Voronoi partition based method is advised to establish the neighborhoods of connected components. We further present a discriminative training algorithm based on the maximum–minimum similarity (MMS) criterion to estimate the parameters in the proposed text extraction approach. Experimental results in Chinese and English text extraction demonstrate the effectiveness of our approach trained with the MMS algorithm, which achieved the precision rate of 93.56% and the recall rate of 98.55% for the test data set. In the experiments, we also show that the MMS provides significant improvement of overall performance, compared with influential training criterions of the maximum likelihood (ML) and the maximum classification error (MCE).     

An information fusion approach to integrate image annotation and text mining methods for geographic knowledge discovery

Due to the steady increase in the number of heterogeneous types of location information on the internet, it is hard to organize a complete overview of the geospatial information for the tasks of knowledge acquisition related to specific geographic locations. The text- and photo-types of geographical dataset contain numerous location data, such as location-based tourism information, therefore defining high dimensional spaces of attributes that are highly correlated. In this work, we utilized text- and photo-types of location information with a novel approach of information fusion that exploits effective image annotation and location based text-mining approaches to enhance identification of geographic location and spatial cognition. In this paper, we describe our feature extraction methods to annotating images, and utilizing text mining approach to analyze images and texts simultaneously, in order to carry out geospatial text mining and image classification tasks. Subsequently, photo-images and textual documents are projected to a unified feature space, in order to generate a co-constructed semantic space for information fusion. Also, we employed text mining approaches to classify documents into various categories based upon their geospatial features, with the aims to discovering relationships between documents and geographical zones. The experimental results show that the proposed method can effectively enhance the tasks of location based knowledge discovery.     

顾及目标关联的自然场景文本检测

目的 目前基于卷积神经网络（CNN）的文本检测方法对自然场景中小尺度文本的定位非常困难。但自然场景图像中文本目标与其他目标存在很强的关联性,即自然场景中的文本通常伴随特定物体如广告牌、路牌等同时出现,基于此本文提出了一种顾及目标关联的级联CNN自然场景文本检测方法。方法 首先利用CNN检测文本目标及包含文本的关联物体目标,得到文本候选框及包含文本的关联物体候选框;再扩大包含文本的关联物体候选框区域,并从原始图像中裁剪,然后以该裁剪图像作为CNN的输入再精确检测文本候选框;最后采用非极大值抑制方法融合上述两步生成的文本候选框,得到文本检测结果。结果 本文方法能够有效地检测小尺度文本,在ICDAR-2013数据集上召回率、准确率和F值分别为0.817、0.880和0.847。结论 本文方法顾及自然场景中文本目标与包含文本的物体目标的强关联性,提高了自然场景图像中小尺度文本检测的召回率。     

静态图像提取文字技术

从图像中提取文字属于信息智能化处理的前沿课题,是当前人工智能与模式识别领域中的研究热点。由于文字具有高级语义特征,对图片内容的理解、索引、检索具有重要作用,因此,研究图片文字提取具有重要的实际意义。又由于静态图像文字提取是动态图像文字提取的基础,故着重介绍了静态图像文字提取技术,总结了几种已提出的算法,并利用计算机语言学方法对提取出的文字进行后期处理,大大提高了文字提取的正确率。     

文本检测与识别在细粒度图片分类中的应用

细粒度的图片分类是深度学习图片分类领域中的一个重要分支,其分类任务比一般的图片分类要困难,因为很多不同分类图片中的特征相似度极高,没有特别鲜明的特征用以区分,因而需要优化一个传统的图片分类方法.在一般的图片分类中,通常通过提取视觉以及像素级别的特征用来训练,然而直接应用到细粒度分类上并不太适配,效果仍有待提高,可考虑利用非像素级别的特征来加以区分.因此,我们提出联合文本信息和视觉信息作用于图片分类中,充分利用图片上的特征,将文本检测与识别算法和通用的图片分类方法结合,应用于细粒度图片分类中,在Con-text数据集上的实验结果表明我们提出的算法得到的准确率有显著的提升.     

中文文本中抽取特征信息的区域与技术

本文探讨了各种从中文文本中抽取特征信息的区域和技术。本文以新闻语料、科技论文、公文类文献为例,详细论述了从各类文本中抽取特征信息的区域与技术,对科技论文,还给出了一些可操作的产生式规则。无论对自动标引、自动分类,还是自动文摘的研究者而言,本文的方法与结论都有一定的参考价值。     

Multitemporal censusing of a population of eastern hemlock (Tsuga canadensis L.) from remotely sensed imagery using an automated segmentation and reconciliation procedure

Large-scale (1:3000) color aerial images of a population of eastern hemlock (Tsuga canadensis L.) were collected in the early spring of 1997, 1998, and 1999. An automated spatial segmentation procedure was developed to identify and measure individual population objects or blobs within the forest population. To ensure the comparability of multiyear segmentation maps, an automated blob reconciliation procedure was also developed to make certain that no hemlock pixels were assigned to different blobs in different years. The automated segmentation and reconciliation procedures were applied to a population of naturally occurring hemlock. Following spatial segmentation, a large majority of hemlock blobs (∼66-71%) were found to be closely associated with ground referenced, manually delineated individual hemlock crowns. The remaining blobs consisted of spatially distinct parts of a crown or closely clumped multiple crowns. Similar overall classification accuracies (∼63-72%) were found following the reconciliation of multitemporal image pairs. The development of these spatially explicit multitemporal population data sets should prove useful to further investigations of the dynamics of and environmental influence on plant populations.     

静态图像提取文字技术

从图像中提取文字属于信息智能化处理的前沿课题,是当前人工智能与模式识别领域中的研究热点。由于文字具有高级语义特征,对图片内容的理解、索引、检索具有重要作用,因此,研究图片文字提取具有重要的实际意义。又由于静态图像文字提取是动态图像文字提取的基础,故着重介绍了静态图像文字提取技术,总结了几种已提出的算法,并利用计算机语言学方法对提取出的文字进行后期处理,大大提高了文字提取的正确率。     

面向荧光显微图像的斑点检测

实体肿瘤学中, 利用荧光原位杂交(FISH)技术处理后的间期细胞核荧光显微图像上, DNA扩增往往呈现为衍射极限斑点, 成像条件限制了图像质量, 导致图像信噪比较低、背景干扰严重且存在非斑点结构干扰. 设计适用的斑点检测方法, 提供客观且定量的数据, 有助于医生对于癌症病情的诊断. 算法首先采用3层小波多尺度求和对荧光图像去噪, 随后利用多尺度高斯拉普拉斯算子增强斑点区域, 最后通过4个方向的单边二阶高斯核抑制非斑点区域, 完成斑点检测. 实验结果表明, 对于自建数据库中83张图像, 算法平均F分数达到0.96, 平均运行时间0.5 s以下.