视频帧内运动目标复制-粘贴篡改检测算法

摘要： 提出了一种基于Lucas_Kanade(LK)光流及运动目标预检机制的视频帧内运动目标复制 粘贴篡改检测算法.该算法分为运动目标检测与跟踪、运动序列筛选和空间域匹配3个阶段.运动目标检测与跟踪利用背景建模算法和卡尔曼滤波器进行检测和跟踪;运动序列的筛选采用LK方法得到各运动序列的光流值,并计算其相关性来选择可能存在篡改的视频帧序列;空间域匹配利用尺度不变特征变换算法对上一阶段得到的对应运动序列逐帧进行匹配,过滤正常的视频序列.实验结果表明,本文算法能有效检测同源视频中针对运动目标的多帧复制 粘贴篡改.     

Segregation of object and background motion in the retina

An important task in vision is to detect objects moving within a stationary scene. During normal viewing this is complicated by the presence of eye movements that continually scan the image across the retina, even during fixation. To detect moving objects, the brain must distinguish local motion within the scene from the global retinal image drift due to fixational eye movements. We have found that this process begins in the retina: a subset of retinal ganglion cells responds to motion in the receptive field centre, but only if the wider surround moves with a different trajectory. This selectivity for differential motion is independent of direction, and can be explained by a model of retinal circuitry that invokes pooling over nonlinear interneurons. The suppression by global image motion is probably mediated by polyaxonal, wide-field amacrine cells with transient responses. We show how a population of ganglion cells selective for differential motion can rapidly flag moving objects, and even segregate multiple moving objects.     

Temporal hyperacuity in the electric sense of fish

It has recently become evident that sensory thresholds for certain tasks are lower than those expected from the properties of individual receptors. This perceptual capacity, termed hyperacuity, reveals the impressive information-processing abilities of the central nervous system. Although much is known about spatial hyperacuity, temporal hyperacuity has received little attention. Here we demonstrate that an electric fish, Eigenmannia, can detect modulations in the timing (phase) of an electrical signal at least as small as 400 ns. Such sensitivity exceeds the temporal resolution of individual phase-coding afferents. This hyperacuity results from a nonlinear convergence of parallel afferent inputs to the central nervous system; subthreshold inputs from particular areas of the body surface accumulate to permit the detection of these extremely small temporal modulations.     

Vernier acuity of neurones in cat visual cortex

The ability of human observers to detect Vernier breaks of as little as 5 s arc has been termed hyperacuity as this distance is substantially less than the angular separation of the bars of the highest spatial frequency of grating (approximately 1 arc min) that can be detected. Although the visual cortex is a likely candidate for the location of detectors involved in this performance, it is not known whether there are cells sensitive enough to detect deviations from co-linearity that are small compared with their spatial resolution (defined in terms of the highest spatial frequency that the cell can detect). We report here the results of physiological experiments on single units in area 17 of the cat visual cortex in which we studied the effect of introducing a Vernier break into a bar stimulus moved across the receptive field of the cell at a constant velocity. Our results show that the responses of most simple and complex cells are significantly reduced by the introduction of a Vernier break that is substantially smaller than the spatial resolution of the cell. The most sensitive cells in our sample could discriminate Vernier offsets of 3-6 arc min with a reliability of approximately 70%. This was much smaller than their spatial resolution, which was in the range 25-30 arc min. We interpret these results in terms of mechanisms that could underly the orientation selectivity of cortical neurones and suggest how our results relate to human Vernier acuity.     

单幅运动模糊图像的恢复及应用

运动模糊图像复原技术高度依赖于点扩散函数的估计,即估计线性运动模糊的方向和长度,这两个参数与物体实际运动速度有一定几何关系。本文在去除运动模糊的同时,提出一种基于单幅模糊图像估计车速的方法。由于运动模糊通常发生在图像的动态区域,因此在恢复图像的同时依据成像几何、相机摆姿和估计出的模糊参数经过数学推导可以计算出车速。本文依据模糊像素扩散现象及模糊图像在频域的性质来估计模糊参数,结合维纳滤波方法进行复原。从而辨识车辆、估计车速、辅助交通管理。实验表明本文方法实用性强,效果较好。     

运动目标的快速检测、跟踪和判别

为完成自然环境中大范围的环境监控 ,实现了一个运动目标检测、跟踪和判别系统。该系统利用一个固定平台上的、有 3 60°旋转和一定俯仰的两自由度摄像机监视自然环境 ,利用 2 -D仿射模型和鲁棒参数估计的主运动分析得到背景运动参数 ,能够在短时间内完成 3 60°全景图的拼接 ,并能利用出格点检测和聚类自动检测、通过限制搜索范围的检测和维护运动目标缓冲池主动跟踪运动目标 ,还能按目标区域的周期性变化判别目标种类 (人或车辆 )。实验表明 ,系统能够实时可靠地检测、跟踪运动目标并完成判别 ,满足特定的监控要求。另外 ,该运动目标判别方法简单可靠 ,其结果可作为视频序列识别和检索的一项重要特征     

混合交通场景中的车辆检测和识别

介绍了一种城市交通流量视频检测系统及其车辆识别新算法。通过检测道路上的运动目标并利用目标的运动信息和外形特征进行车辆识别。考虑到目标的阴影会影响目标的形状特征 ,导致目标识别错误 ,提出了一种新的阴影检测算法———利用目标的灰度特征对运动目标进行阴影检测和分离。实验结果表明该系统在模拟城市混合交通环境下 ,能够克服目标阴影的影响 ,准确检测和识别车辆 ,同时能够满足实时性的要求     

基于运动平台的运动目标检测与跟踪

为了实现大视场智能监控,本文提出了一种基于运动平台的运动目标检测与跟踪方法.在相邻帧之间通过块匹配进行运动补偿,采用三帧差分法分割出运动目标.当运动目标正常运动时跟踪其形心;当运动目标被遮档时,根据卡尔曼滤波器预测的形心跟踪目标.其中,对于块匹配,采用边缘点作为匹配块的中心点并根据摄像机的运动方向确定搜索范围,使处理速度提高了38.6%.另外,比较得出最小二乘法对运动目标运动状态突然改变时拟合效果差,因此采用卡尔曼滤波器进行预测.实验证明,本算法适应环境变化的能力强,而且平均每秒处理37.6帧,达到实时处理要求.     

fMRI evidence for objects as the units of attentional selection.

Contrasting theories of visual attention emphasize selection by spatial location, visual features (such as motion or colour) or whole objects. Here we used functional magnetic resonance imaging (fMRI) to test key predictions of the object-based theory, which proposes that pre-attentive mechanisms segment the visual array into discrete objects, groups, or surfaces, which serve as targets for visual attention. Subjects viewed stimuli consisting of a face transparently superimposed on a house, with one moving and the other stationary. In different conditions, subjects attended to the face, the house or the motion. The magnetic resonance signal from each subject's fusiform face area, parahippocampal place area and area MT/MST provided a measure of the processing of faces, houses and visual motion, respectively. Although all three attributes occupied the same location, attending to one attribute of an object (such as the motion of a moving face) enhanced the neural representation not only of that attribute but also of the other attribute of the same object (for example, the face), compared with attributes of the other object (for example, the house). These results cannot be explained by models in which attention selects locations or features, and provide physiological evidence that whole objects are selected even when only one visual attribute is relevant.     

Spatial filtering precedes motion detection.

When we perceive motion on a television or cinema screen, there must be some process that allows us to track moving objects over time: if not, the result would be a conflicting mass of motion signals in all directions. A possible mechanism, suggested by studies of motion displacement in spatially random patterns, is that low-level motion detectors have a limited spatial range, which ensures that they tend to be stimulated over time by the same object. This model predicts that the direction of displacement of random patterns cannot be detected reliably above a critical absolute displacement value (Dmax) that is independent of the size or density of elements in the display. It has been inferred that Dmax is a measure of the size of motion detectors in the visual pathway. Other studies, however, have shown that Dmax increases with element size, in which case the most likely interpretation is that Dmax depends on the probability of false matches between pattern elements following a displacement. These conflicting accounts are reconciled here by showing that Dmax is indeed determined by the spacing between the elements in the pattern, but only after fine detail has been removed by a physiological prefiltering stage: the filter required to explain the data has a similar size to the receptive field of neurons in the primate magnocellular pathway. The model explains why Dmax can be increased by removing high spatial frequencies from random patterns, and simplifies our view of early motion detection.     

基于分区灰度投影稳像的运动目标检测算法

针对视频监控系统中,复杂环境引起摄像机抖动,造成运动目标检测不准确的问题,提出了一种基于分区灰度投影稳像的运动目标检测算法.首先对每帧图像进行分区,利用分区灰度投影算法对图像各分区的运动矢量进行准确提取和相关性分析,进行抖动判断,并对抖动帧进行运动补偿.然后利用高斯混合背景建模算法进行运动目标提取.最后对目标提取结果进行形态学处理,以进一步提高目标提取的精度.实验结果表明,本文算法较好地消除了场景中运动目标对运动矢量计算的干扰,实现了在摄像机抖动视频场景中的运动目标的准确检测和提取,大大降低了抖动视频目标检测的虚警率.     

Transparency and coherence in human motion perception

When confronted with moving images, the visual system often must decide whether the motion signals arise from a single object or from multiple objects. A special case of this problem arises when two independently moving gratings are superimposed. The gratings tend to cohere and move unambiguously in a single direction (pattern motion) instead of moving independently (component motion). Here we report that the tendency to see pattern motion depends very strongly on the luminance of the intersections (that is, to regions where the gratings overlap) relative to that of the gratings in a way that closely parallels the physics of transparency. When the luminance of these regions is chosen appropriately, pattern motion is destroyed and replaced by the appearance of two transparent gratings moving independently. The observations imply that motion detecting mechanisms in the visual system must have access to tacit 'knowledge' of the physics of transparency and that this knowledge can be used to segment the scene into different objects. The same knowledge could, in principle, be used to avoid confusing shadows with real object boundaries.     

Self-motion and the perception of stationary objects

One of the ways that we perceive shape is through seeing motion. Visual motion may be actively generated (for example, in locomotion), or passively observed. In the study of the perception of three-dimensional structure from motion, the non-moving, passive observer in an environment of moving rigid objects has been used as a substitute for an active observer moving in an environment of stationary objects; this 'rigidity hypothesis' has played a central role in computational and experimental studies of structure from motion. Here we show that this is not an adequate substitution because active and passive observers can perceive three-dimensional structure differently, despite experiencing the same visual stimulus: active observers' perception of three-dimensional structure depends on extraretinal information about their own movements. The visual system thus treats objects that are stationary (in an allocentric, earth-fixed reference frame) differently from objects that are merely rigid. These results show that action makes an important contribution to depth perception, and argue for a revision of the rigidity hypothesis to incorporate the special case of stationary objects.     

移动目标的自动监视报警方法

为了快速准确地从背景中检测出移动目标,并控制云台对目标进行自动跟踪,提出了一种移动目标的自动监视报警方法。该方法充分利用连续图像特性,采用改进的帧间减影法检测运动,并用差分图像的阈值化方法分割出动目标;进而采用光谱滤波等技术提高图像传感器的探测灵敏度,进一步提高了该方法的作用距离。仿真结果表明,该方法可以有效、准确地对移动目标进行检测和跟踪。     

具有前景目标的动态场景视频快速稳像算法

为了提高具有前景目标的动态场景视频的稳像效果,采用了一种基于块绝对差分的运动目标自动提取方法。在提取运动目标的基础上,提出通过自适应设定相关阈值的方法,来预判相关背景模块是否具有足够的信息,以减少用于运动矢量估计的模块数量。仿真实验结果表明,对于具有前景目标的动态场景视频,在减少运动目标提取时间的基础上,能够有效提高稳像算法的速度和精度,对实际视频取得了比较理想的稳像效果。     

Motion-induced spatial conflict

Borders defined by small changes in brightness (luminance contrast) or by differences in colour (chromatic contrast) appear to move more slowly than those defined by strong luminance contrast. As spatial coding is influenced by motion, if placed in close proximity, the different types of moving border might appear to drift apart. Using this configuration, we show here that observers instead report a clear illusory spatial jitter of the low-luminance-contrast boundary. This visible interaction between motion and spatial-position coding occurred at a characteristic rate (approximately 22.3 Hz), although the stimulus motion was continuous and invariant. The jitter rate did not vary with the speed of movement. The jitter was not due to small involuntary movements of the eyes, because it only occurred at a specific point within the stimulus, the low-luminance-contrast boundary. These findings show that the human visual system contains a neural mechanism that periodically resolves the spatial conflict created by adjacent moving borders that have the same physical but different perceptual speeds.     

一种固定背景下运动视频对象的自动分割方法

提出了一种固定背景运动视频对象的自动分割方法．该方法通过互帧差四次矩检测运动对象，利用背景减除对点分类，最后用彩色分水线算法得到运动对象实验证明该方法能够对固定背景下运动视频对象进行实时自动分割．     

基于DM642的自适应运动目标检测系统

针对帧差法和背景差法的运动检测方法的不足以及运动检测系统中图象数据处理量大、算法复杂等特点,设计了以TMS320DM642为核心处理器运动目标检测系统硬件平台。运动目标检测算法采用三帧差法与背景差法相结合的方法,并使用一种类单高斯阈值求解的方法实现门限的自适应调整和运动目标的分割,弥补了帧差法和背景差法的缺陷,提高了检测的准确性。结果表明,该系统运行良好,能够准确地检测出运动目标。     

动态场景中运动目标检测与跟踪

为了在静态和动态场景中均能实现对运动目标的检测与跟踪,提出了基于运动检测和视频跟踪相结合的视频监控方法. 建立四参数运动仿射模型来描述全局运动,采用块匹配法对其进行参数估计;采用基于全局运动补偿的Horn-Schunck算法检测出运动目标;使用卡尔曼滤波对运动目标的质心位置、宽度和高度进行跟踪. 实验结果表明,该方法能够有效地对静态和动态场景中运动目标进行检测与跟踪.     

Temporal dynamics of a neural solution to the aperture problem in visual area MT of macaque brain

A critical step in the interpretation of the visual world is the integration of the various local motion signals generated by moving objects. This process is complicated by the fact that local velocity measurements can differ depending on contour orientation and spatial position. Specifically, any local motion detector can measure only the component of motion perpendicular to a contour that extends beyond its field of view. This "aperture problem" is particularly relevant to direction-selective neurons early in the visual pathways, where small receptive fields permit only a limited view of a moving object. Here we show that neurons in the middle temporal visual area (known as MT or V5) of the macaque brain reveal a dynamic solution to the aperture problem. MT neurons initially respond primarily to the component of motion perpendicular to a contour's orientation, but over a period of approximately 60 ms the responses gradually shift to encode the true stimulus direction, regardless of orientation. We also report a behavioural correlate of these neural responses: the initial velocity of pursuit eye movements deviates in a direction perpendicular to local contour orientation, suggesting that the earliest neural responses influence the oculomotor response.