共查询到20条相似文献,搜索用时 122 毫秒
1.
超分辨率复原技术是一种可用于提高图像细节辨识能力的有效方法。其在视频监控领域可望得到广泛应用。超分辨率图像处理技术通过融合多帧相似的低分辨率图像达到提高图像细节的目的。从而降低对监控视频采集硬件与后端辅助处理系统的要求,提高对特定目标的辨析能力。本文重点介绍了在视频监控领域较为实用的凸集投影算法、最大后验概率估计算法、基于对象的超分辨率复原方法、基于示例学习与多类预测器的超分辨率复原方法。对以上超分辨率复原方法实现流程的优缺点与其在视频图像监控领域的应用方法进行了相应分析。分析了超分辨率视频监控图像复原常用的基于块匹配与光流的对象运动估计方法。对超分辨率复原重建图像质量的评估标准也进行了相应讨论。 相似文献
2.
Snake算法能够跟踪运动图像中对象的非刚性运动,但是对于背景复杂的图像,Snake跟踪的结果不够理想。因而在首帧分割得到对象轮廓的二值模型后,再采用基于Hausdorff距离的跟踪器,找到对象模型在后继帧中的最佳匹配位置;然后采用Snake模型对该匹配位置上的非刚性形变的像素进行匹配。实验表明:对于具有静止背景且前景对象不是快速运动的视频序列,与直接采用Snake技术进行运动对象的跟踪相比,该提取视频对象平面过程能够进一步提高结果的正确性。 相似文献
3.
一种用于动态视频超分辨率的多尺度最小二乘仿射块匹配图像配准方法 总被引:2,自引:0,他引:2
动态视频的超分辨率复原中,连续各帧图像间的精确匹配具有非常重要的意义。该文提出一种基于多尺度最小二乘仿射块匹配的图像配准方法。首先定义了一个指标Dmv来衡量图像的整体和局部匹配效果,并以此为基础设计了一种多尺度块选择机制,根据图像的运动情况选择匹配块大小,以兼顾图像中运动平坦和非平坦区域的匹配效果。与传统的块匹配方法不同,该文采用基于仿射模型的最小二乘配准方法实现各图像块的匹配,并通过修正步长的归一化处理解决了不同大小图像块在匹配时的收敛问题,从而在提高参数估计精度的同时降低了算法的运算量。最后,通过实验对算法的匹配性能及其对超分辨率复原算法整体性能的影响进行了测试。实验结果表明,该方法不仅可以实现更为准确的运动估计,当用于最大后验概率MAP超分辨率复原算法时,能够进一步有效提高算法的复原性能和实现速度。 相似文献
4.
5.
基于多个非刚体目标跟踪的视频对象平面生成算法 总被引:1,自引:0,他引:1
提出了一种提取运动对象的新的视频序列分割算法。算法的核心是一个对象跟踪器,它利用一种基于对象行为的跟踪算法对多个非刚体目标有效地进行对象跟踪,并与后续帧进行匹配,然后采用一种基于运动相连成分的模型刷新方法对模型的每一帧进行刷新,初始的模型自动产生,再利用滤波技术滤除静止背景,最后,利用边界图像模型从序列中提取出视频对象平面(VOP)。 相似文献
6.
7.
为了提高视频的空间分辨率,提出了一种利用帧间运动信息进行超分辨率重建的方法。对于整个视频的重建,提出了一种基于滑动窗的分段重建模型。在每一个滑动窗中,首先对相邻帧进行子像素级精度的运动配准;然后通过迭代反投影算法进行超分辨率重建。在配准算法中,提出了一种基于四参数刚体变换模型的配准方法,通过迭代求解和高斯金字塔图像模型由粗及精地进行运动估计。分别对模拟图像及实拍彩色视频进行重建,实验结果表明,该配准算法具有较高的精度,重建算法取得了较高的峰值信噪比(PSNR)值,重建视频具有更好的视觉效果和更高的分辨率能力,可被广泛应用于在帧间主要存在平移和旋转运动的视频序列的超分辨率重建。 相似文献
8.
针对目标探测器在大气中高速飞行时受湍流干扰,导致光学系统接收到的视频/图像产生像素偏移、模糊、信噪比降低等问题,本文对湍流退化视频/图像复原的复杂性及复原方法进行了研究,提出了一种基于非凸势函数优化与动态自适应滤波的湍流退化视频复原方法。首先,研究了湍流退化视频的求和与去模糊框架,并通过利用非刚性配准方法对刚性全局配准方法进行改进,进一步缩小了模糊核的尺度;然后,在计算机视觉的非凸优化框架下,构建了图像解卷积的非凸性算法,有效地解决了图像解卷积难题;最后,结合湍流退化视频自身特点,对超分辨率视频复原的动态自适应滤波框架进行了扩展与改进,使其适用于湍流退化视频的复原。仿真实验结果表明,本文方法的复原效果不仅有较大提升,而且实现了对湍流退化视频序列的动态自适应复原。 相似文献
9.
10.
为实现对车载设备视频图像中车辆的识别和跟踪,针对图像中的运动目标和动态背景,提出了一种基于特征学习的目标检测和超像素跟踪算法.该算法首先对训练图像进行HOG特征提取,并利用AdaBoost算法得到强分类器.利用强分类器对采集的图像进行车辆检测,从而确定搜索区域.结合对搜索区域的超像素分割结果,采用均值漂移聚类算法实现车辆识别与跟踪.实验结果表明,该算法可以很好地实现视频序列中的车辆识别,提高了目标跟踪的实时性. 相似文献
11.
Wang Suyu Shen Lansun 《电子科学学刊(英文版)》2008,25(1):140-144
Construction of high resolution images from low resolution sequences is often important in surveillance applications. In this letter, an affine based multi-scale block-matching image registration algorithm is first proposed. The images to be registered are divided into overlapped blocks of different size according to its motions. The Least Square (LS) image registration algorithm is extended to match the blocks. Then an object based Super Resolution (SR) scheme is designed, the Maximum A Priori (MAP) super resolution algorithm is extended to enhance the resolution of the interest objects. Experimental results show that the proposed multi-scale registration method provides more accurate registration between frames. Further more, the object based super resolution scheme shows an enhanced performance compared with the traditional MAP method. 相似文献
12.
A practical technique is developed to determine the electric and/or magnetic field on objects and sources inside a spherical measurement surface. The technique, known as spherical microwave holography (SMH), provides a nondestructive, nonintrusive method of point-by-point evaluation of antennas and radomes over their spatial extent. The resolution capability of SMH is developed and demonstrated by measurements. Resolution in SMH is only limited by the measurement system's capabilities. Dielectric and metallic obstacles on the surface of a radome are located and identified. Resolution as small as 0.33λ0 is demonstrated 相似文献
13.
14.
In this work, we bring together object tracking and digital watermarking to solve the spatio-temporal object adjacency problem
in image sequences. Spatio-temporal relationships are established by embedding objects with unique digital watermarks and
then by propagating the watermark frame by frame. Watermark propagation is accomplished by an existing object tracking module
so that a tracked object acquires its watermark from the correspondences established by the object tracker. The spatio-temporally
marked image sequences can then be searched to establish spatial and temporal adjacency among objects without using traditional
spatio-temporal graphs. Borrowing from graph theory, we construct binary adjacency matrices among tracked objects and develop
interpretation rules to establish a track history for each object. Track history can be used to determine the arrival of new
objects in frames or the changing of spatial and temporal positions of objects with respect to each other as they move through
time and space. 相似文献
15.
Many videos capture and follow salient objects in a scene. Detecting such salient objects is thus of great interests to video analytics and search. However, the discovery of salient objects in an unsupervised way is a challenging problem as there is no prior knowledge of the salient objects provided. Different from existing salient object detection methods, we propose to detect and track salient object by finding a spatio-temporal path which has the largest accumulated saliency density in the video. Inspired by the observation that salient video objects usually appear in consecutive frames, we leverage the motion coherence of videos into the path discovery and make the salient object detection more robust. Without any prior knowledge of the salient objects, our method can detect salient objects of various shapes and sizes, and is able to handle noisy saliency maps and moving cameras. Experimental results on two public datasets validate the effectiveness of the proposed method in both qualitative and quantitative terms. Comparisons with the state-of-the-art methods further demonstrate the superiority of our method on salient object detection in videos. 相似文献
16.
This paper proposes a new method to describe and identify a 3-D curved object for the purpose of validating a fabricated object to the design specification. Curved 3-D objects are, in general, difficult to represent and identify because they lack distinct properties such as edges, planes, or cylindrical surfaces which are the building blocks in representing objects. In this paper, the authors propose to use principal axes of a 3-D object to establish a reference for the representation. A method of obtaining an inertia matrix from a 3-D range image is developed. The unique set of principal axes is obtained from the inertia matrix of an object with an arbitrary 3-D position and orientation, and the object can be described uniquely on these principal axes. On the principal axes, an object is described by a set of features describing the shape of the object such as spine, section size, section orientation, and section contraction. The features are used for comparing two objects for the validation purpose. The authors also propose a direct measure of similarity between two objects as a mean-squared difference of radii. As an experiment, two 3-D object models are designed through a CAD package, and fabricated objects are compared with the designed models for validation purposes 相似文献
17.
Chen Tingbiao 《中国邮电高校学报(英文版)》1996,(1)
AMethodfor3DSceneDescriptionandSegmentationinanObjectRecord¥ChenTingbiao(DepartmentofRadioEngineering,NamingUniversityofPosts... 相似文献
18.
对象缓存是一种通过在使用对象后不立即释放,而是存储在内存或硬盘中并被后来的客户端请求重用,避免重新建立对象的昂贵成本的机制。在考查了业界广泛使用的几种对象缓存框架后,提出了一种分布式对象缓存框架的设计方案LiteCS。该框架中服务器不需要负责对象更改消息的传递,即使缓存对象被频繁地修改,也不会大大增加系统的整体负载,可有效降低网络负荷.现所做的两组测试证明了LiteCS框架适用于多台服务器通过网络共享缓存对象。 相似文献
19.
20.
SAR图像多尺度目标检测能够实现大场景SAR图像中关键目标的定位与识别,是SAR图像解译的关键技术之一。然而针对尺寸相差较大的SAR目标的同时检测,即跨尺度目标检测问题,现有目标检测方法难以实现。该文提出一种基于特征转移金字塔网络(FTPN)的SAR图像跨尺度目标检测方法。在特征提取阶段采用特征转移方法,实现各层特征图的有效连接,实现不同尺度特征图的提取;同时采用空洞卷积群方法,增大特征提取的感受野,促使网络提取到大尺度目标特征。上述环节能够有效保留不同尺寸目标特征,从而实现SAR图像中跨尺度目标的同时检测。基于高分三号SAR数据、SSDD数据集及高分辨率SAR舰船检测数据集-2.0等数据集的试验表明,该文方法能够实现SAR图像中机场、舰船等跨尺度目标的检测,在已有数据集上mAP达96.5%,较特征金字塔网络算法提升8.1%,并且整体性能优于现阶段最新的YOLOv4等目标检测算法。 相似文献