Flexible and Robust Model Matching based on Association Graph for Form Image Understanding

A new method of image understanding for forms based on model matching is proposed in this paper as the basis of an OCR which can read a variety of forms. The outline of this method is described as follows. First, ruled lines are extracted from the input image of a form. After that, several lines are grouped as one to be recognised as data corresponding to a sub-form. These lines and sub-forms are both used for understanding the form, taking into account their feature attributes and the relationships between them. Each feature in the input image of a form is expected to correspond to a feature in one of the model forms, which are described as structured features. This correspondence is represented by a node in an association graph, where an arc represents compatible correspondences established on the basis of feature relationships. The best match is found as the largest maximal clique in the association graph. Experimental results show the method is robust and effective for document images of poor quality, and also for various styles of forms.     

Preprocessing of Low-Quality Handwritten Documents Using Markov Random Fields

This paper presents a statistical approach to the preprocessing of degraded handwritten forms including the steps of binarization and form line removal. The degraded image is modeled by a Markov Random Field (MRF) where the hidden-layer prior probability is learned from a training set of high-quality binarized images and the observation probability density is learned on-the-fly from the gray-level histogram of the input image. We have modified the MRF model to drop the preprinted ruling lines from the image. We use the patch-based topology of the MRF and Belief Propagation (BP) for efficiency in processing. To further improve the processing speed, we prune unlikely solutions from the search space while solving the MRF. Experimental results show higher accuracy on two data sets of degraded handwritten images than previously used methods.     

基于多传感器的全向足球机器人自定位

随着对足球机器人智能水平的要求进一步提高,机器人足球委员会将比赛场地的立柱、球门颜色取消.这使以前的基于球门、立柱地标的足球机器人自定位方法失效了.本文提出了一种利用了里程计、罗盘和全景摄像头多种传感器信息的视觉图像特征匹配的足球机器人自定位方法.首先,机器人通过里程计和罗盘取定一个可能位姿.然后,由视觉处理系统把机器人的可能位姿当作变换因子对实时拍摄到的场景图像作旋转、平移变换.最后,将变换后的图像中的白线与参考图像中的白线相比较,选择使图像匹配程度最大的变换因子作为机器人自定位的结果.实验结果表明该自定位方法达到了较高定位精度并能满足比赛的高实时性要求.     

Application of Super High Definition Images in Telemedicine: System Requirements and Technologies for Teleradiology and Telepathology

It was recognized early on that the digitization of medical information would advance the efficiency of diagnostic technology. However, the digitization of image data, which makes up the majority of medical information, is dependent on advances in technologies such as input, processing, transmission, storage, and display. Insufficient advances in such technologies has effectively limited the digitization of image data for medical use. The result of this has been non-networked systems or LANs confined to a single hospital. Such isolated systems integrate only portions of digital medical images such as x-ray computer tomography (CT), magnetic resonance (MR), and computed radiography (CR).Fortunately, recent advances in the areas of super high definition image I/O, high-quality encoding, super high speed transmission, and high-capacity storage has turned the tide in favor of the digitization and networking of all medical information. This paper will focus on the digitization and networking of medical image information used within hospitals and provide a multifaceted study of the technologies necessary for these advances. This will allow us to discuss the present state of related technical developments and the level that has been attained so far. In addition, we have targeted image information that demands the highest level of quality (radiological and pathological images) for application in medical diagnosis using super high definition images, the image technology being developed by the authors of this paper. We will cover the concrete issues and approaches to solutions that must be investigated when building and networking a digital system.     

Privacy aware image template matching in clouds using ambient data

Cloud computing is ideal for image storage and processing because it provides enormously scalable storage and processing resources at low cost. One of the major drawbacks of cloud computing, however, is the lack of robust mechanisms for the users to control the privacy of the data they farm out to the clouds. In this paper, we develop an image encoding scheme that enhances the privacy of image data that is outsourced to the clouds for processing. Unlike previously proposed image encryption schemes, our encoding scheme allows different forms of pixel-level image processing to take place in the clouds while the actual image is not revealed to the cloud provider. Our encoding scheme uses a chaotic map to transform the image after it is masked with an arbitrarily chosen ambient image. We use template matching as a common image processing task to demonstrate the ability of our scheme to perform computations on privacy enhanced images. A simplified prototype of the image processing system was implemented and the experimental results are presented in this paper.     

基于DSP的JPEG标准编解码系统设计

为了将图像数据高质量地传输,便于图像数据的检索、分析、处理和存储,设计了一个以JPEG标准的静态图像压缩编码系统;将原始图像进行图像压缩编码,同时利用DSP芯片处理速度快的特点,进行核心算法的处理满足系统实时性的要求;经过系统的多次验证,图像压缩编码系统成功地将一个31K大小的BMP格式图像转换为3K大小的JPEG格式图像,完成了10∶1的JPEG标准静态图像压缩;因此将原始图像进行图像压缩编码是解决图像数据量大小与有限存储容量和传输带宽矛盾的最合理方法。     

Hadoop平台下新型图像并行处理模型设计

Hadoop在处理海量小图像数据时，存在输入分片过多以及海量小图像存储问题。针对这些问题，不同于采用HIPI、SequenceFile等方法，提出了一个新型图像并行处理模型。利用Hadoop适合处理纯文本数据的特性，本模型使用存储了图像路径的文本文件替换图像数据作为输入，不需要设计图像数据类型。在Map阶段直接完成图像的读取、处理、存储过程。为了简化图像处理算法，将OpenCV和Map函数结合并设计了对应的存储方法，实现小图像文件的存储。实验表明，在Hadoop分布式系统平台下，模型不论在小数据量还是在大数据量的测试数据环境中，都具有良好的吞吐性能和稳定性。     

新颖的频域图像超分辨重建方法

图像频域解混叠超分辨算法要求输入低分辨率图像帧数及其帧间亚像元位移满足一定的条件,因此限制了其应用范围。利用单帧超分辨技术和再采样函数从每帧输入低分辨率图像产生16帧相同分辨率的图像,再从中挑选满足限制条件的图像帧,从而消除了原算法要求的两个主要限制条件,建立了改进的二至多帧输入图像频域解混叠超分辨算法。三组仿真实验结果表明,该算法消除了输入帧数和帧间亚像元位移的限制条件,并且可以使峰值信噪比提高5dB左右。     

卫星遥感图像并行几何校正算法研究

几何校正是遥感图像处理过程中的重要环节，具有计算量大、耗时长的特点，导致遥感图像处理的效率低下．该文提出一种分布存储环境下的并行几何校正算法，每个处理器通过计算本地输入子图像在目标图像中的范围，确定其需要进行重采样计算的区域，使计算过程中所需的数据均为本地数据，很好地解决了数据局部性问题．文章利用首尾相连的闭线段近似表示理想的输出图像块边界这一思想，详细讨论了局部输出区域的计算方法，并采用一种新的存储结构用于保存校正后的输出图像块信息．在机群系统上对算法进行实现，结果表明该算法具有良好的并行性能．     

Structured light system-based selective data acquisition

Selective data acquisition has been introduced recently for direct interfacing reverse engineering (RE) with rapid prototyping (RP). Data are acquired selectively and locally layer by layer according to the analysis of an image processing algorithm that has been developed for recommending the scanning positions from the edge images of the object. The recommendations, however, depend on the quality of the input images and the edge detection parameters. An attempt to make selective data acquisition more robust and less sensitive to the aforementioned factors is therefore presented in this paper. A structured light system (SLS) and projective transformation have been integrated into this SLS-based selective data acquisition to induce features on an object surface to appear explicitly through phase distortion of a projected bright line, and to diminish perspective distortion effects in the analysis through transformation of deformed bright lines on side views to be contour segments on top view. The implementation of this new selective data acquisition approach is also presented and discussed.     

基于CPLD芯片控制的视频图像处理系统设计

针对当前连续视频图像的空间冗余性较高,图像数据处理及时性较差的问题,设计了基于CPLD芯片控制的视频图像处理系统。将CPLD存储模块内提取的数据信息,平均分配至视频图像采集元件及图像边缘检测元件中,再借助信息传输信道,建立与VGA视频图像显示模块的物理连接,实现视频图像处理系统的硬件执行方案设计。利用分数像素差优化结果,规范子窗口的动态显示权限,控制连续视频图像在空间范围内的冗余化程度,完成图像源路数的智能控制处理。联合视频进程与多线程控制原理,实施数据包缓冲区的信息排序,再建立全新的视频流存储模式,实现视频流的完整播放。对比实验结果表明,与多路视频处理手段相比,应用CPLD芯片控制处理系统后,图像数据帧缓存速率明显提升,而VGA信息转换指标却大幅下降,在抑制连续视频图像空间冗余能力的同时,解决了图像数据处理不及时的问题。     

基于DSP的无人机遥感影像SIFT算法设计与实现

为了满足尺度不变特征变换（SIFT）算法临场处理大尺寸无人机（UAV）组网遥感观测影像的实时快速需求,提出一种基于数字信号处理器（DSP）内核的硬件乘法器来处理单精度浮点型像素数据乘法的算法实现方案。首先,根据DSP内核的硬件乘法器的数据输入、输出特性,重构SIFT算法的图像数据结构和图像函数,以实现硬件乘法器对SIFT算法单精度浮点型像素数据的乘法计算;其次,采用软件流水技术重新编排迭代计算,以增强算法的并行计算能力;最后,将在算法计算过程中产生的动态数据迁移至第三代双倍速率同步动态随机存储器（DDR3）中,以提升算法数据的存储空间。实验结果表明,DSP平台的SIFT算法可以实现对1 000×750的UAV遥感影像的高精度快速处理,所提方案满足无人机组网遥感影像临场处理对SIFT算法的实时快速要求。     

基于DVI接口的图像总线控制系统

设计了一种基于DVI接口的高速图像采集控制系统。该系统能够稳定采集Camlink接口的高速CCD传输的数字信号,并能提供一路DVI接口高清显示和一路PAL制式复合视频。系统可与上位机通过CAN总线实现数据交互,并提供数字视频信号给压缩存储单元和图像处理单元。系统硬件结构简单,工作稳定可靠,能够广泛应用于图像处理领域。     

基于DSP的图像压缩系统设计与算法研究

提出一种基于DSP和FPGA协同设计实现视频图像压缩的控制逻辑方案。由FPGA模块来实现图像采集,DSP模块进行编码压缩,同时针对块匹配算法中搜索精度与计算复杂度相关性问题,介绍了一种基于块匹配的量子行为的微粒群优化算法（Block Match Quantum-behaved Particle Swarm Optimization,BMQPSO）。在图像的实时压缩算法处理中,先对原始图像序列每一帧的宏块用微粒子进行搜索,再根据收敛性要求对压缩编码进行优化。实验结果表明该算法压缩效果优于经典搜索算法。     

循环生成对抗网络的线稿图像自动提取

目的 动漫制作中线稿绘制与上色耗时费力,为此很多研究致力于动漫制作过程自动化。目前基于数据驱动的自动化研究工作快速发展,但并没有一个公开的线稿数据集可供使用。针对真实线稿图像数据获取困难,以及现有线稿提取方法效果失真等问题,提出基于循环生成对抗网络的线稿图像自动提取模型。方法 模型基于循环生成对抗网络结构,以解决非对称数据训练问题。然后将不同比例的输入图像及其边界图输入到掩码指导卷积单元,以自适应选择网络中间特征。同时为了进一步提升网络提取线稿的效果,提出边界一致性约束损失函数,确保生成结果与输入图像在梯度变化上的一致性。结果 在公开的动漫彩色图像数据集Danbooru2018上,应用本文模型提取的线稿图像相比于现有线稿提取方法,噪声少、线条清晰且接近真实漫画家绘制的线稿图像。实验中邀请30名年龄在2025岁的用户,对本文以及其他4种方法提取的线稿图像进行打分。最终在30组测试样例中,本文方法提取的线稿图像被认为最佳的样例占总样例84%。结论 通过在循环生成对抗网络中引入掩码指导单元,更加合理地提取彩色图像的线稿图像,并通过对已有方法提取效果进行用户打分证明,在动漫线稿图像提取中本文方法优于对比方法。此外,该模型不需要大量真实线稿图像训练数据,实验中仅采集1 000幅左右真实线稿图像。模型不仅为后续动漫绘制与上色研究提供数据支持,同时也为图像边缘提取方法提供了新的解决方案。     

基于形态特征提取的图像匹配搜索技术研究

传统的图像搜索方法一般是由图像处理软件自动抽取图像的颜色、形状、纹理等特征,并以此建立特征索引库,进而由用户输入要查找的物品图像,从而找出与之具有相近特征的图像。而文中给出了从数学形态学的角度来提取图像的关键形态特征,然后建立海量物品图片的形态细化骨架库,并以此简化图像搜索的关键内容,降低数据库存储量,提高匹配效率以及准确性的具体方法。     

刺绣图案中直线和圆弧的矢量化研究

多数刺绣图案的基本组成图元是由线段和圆弧组成。针对它的录入问题,本文提出了一种从BMP图像格式到DXF图形格式转换的实现算法。首先通过扫描仪输入刺绣图案图像;然后利用图像处理和多项式拟和算法来实现图像到图形的矢量化;最后形成DXF文件格式。实验结果表明:该算法可有效地提高输图效率。     

A microprocessor-based still frame capturing system

This paper describes the design and implementation of a low cost image capturing system. The basic feature of the design is the massive reduction of the storage requirement in comparison with a fully buffered system. This has been accomplished by effecting a compromise between the amount of high speed storage and the time required to store all or part of a digitized frame, defined as 512 scan lines, each containing 512 picture elements in the format of eight-bit words. The system can be used for the transmission of pictures, texts or documents, where the need for storing pictorial data before transmission can be totally eliminated. It can also be employed as a low cost input technique for a digital image processing system where the required image can be stored in a computer memory and undergo various image processing operations.     

基于粗糙集理论的图像分割智能决策方法

尽管如今已有多种图像分割算法,但是没有任何一种分割方法能够适用于所有的图像.为了使图像跟踪系统能根据图像特征自适应选取分割算法,给出了一种基于粗糙集理论的图像分割智能决策方法.该方法首先选取若干具代表性的分割算法构成算法库,并用它们对各种样本图像进行分割;然后利用从样本图像中提取出来的各种数值特征,并根据图像分割质量评价标准评判出各样本图像的最优分割算法,用其构成决策信息表;最后应用粗糙集理论来对决策信息表进行离散化处理和属性约简,以生成图像分割算法选取的决策规则.该决策方法解决了图像跟踪系统中分割算法选取的一系列难题.实验证明,该决策方法能比较有效地根据系统所处理图像的特征选取出算法库中最优的分割算法,并可满足车载图像跟踪系统的实时性要求.