基于可变形模板的视频对象分割

提出一种基于原型的可变形模板进行图像分割的算法，从而可以将感兴趣的视频对象从静止的复杂背景中提取出来，首先定义一个原型模板来特征化感光趣的视频对象，得到一个先验概率密度函数的一个将运动信息与边缘方向结合起来的似然函数，然后应用贝叶斯准则将两者结合起来，得到后验概率密度函数。求它的最大值，从而确定对象的位置。     

广角无限视景显示系统后投射屏的设计与研究

根据球面反射镜成像理论,利用改变光轴的方法,推导出了广角无限视景显示系统后投射屏面形的计算公式,完成了后投射屏最佳面形的设计。对后投射屏外表面散射层的材料和厚度进行了探讨,给出了最佳的选择方案。首次通过热吹模具成型法研制了高精度的整体投射屏。通过试验验证,该投射屏满足设计要求。使用该投射屏的广角无限显示系统具有图像清晰、视场角大、纵深感强等优点,同时观察者的视点还可在一定的出瞳范围内移动。     

基于GIS的河道、蓄滞洪区洪水演进动态可视化仿真研究

结合地理信息系统(GIS)技术和三维仿真图示技术,探讨了基于GIS的河道、蓄滞洪区洪水演进可视化仿真技术原理与方法,分析了洪水演进可视化模型的建立过程,设计出河道、蓄滞洪区洪水演进系统的总体结构及开发模式,并介绍了系统具体功能的实现.本仿真模型在实例中得到了成功的应用,可为防洪减灾决策提供快捷、直观形象的信息支持.     

Visual Acuity Depends on Perceived Size

Increasing (or decreasing) the convergence angle can shrink (or expand) the perceived size of an object without changing its retinal size. The present report deals with the question of whether such a change in perceived size affects visual acuity. We investigated the effects of perceived size on the legibility of letters, using a telestereoscope which can control the observer’s convergence angle while keeping the size of the retinal image constant. We demonstrated that letters do become more legible (illegible) as perceived size expands (shrinks), although their retinal size stays constant. Then, we measured the ‘threshold’ size of the convergence angle, at which letters became legible from illegible, for several retinal sizes (0.1 to 0.2 deg) of letters. The result shows that the threshold decreases as the retinal size of letters decreases, so that the minimum retinal size of legible letters depends on the convergence angle. This implicates contributions from the size perception at higher levels of the visual system to the determination of visual acuity.     

LED大屏幕视频图象显示亮度校正的一种方法

在大屏幕视频显示的研究中,通常将侧重点放在显示驱动电路的优化上。本文从人眼的主观视觉理论出发,讨论了大屏幕视频图象显示与主观视觉之间的关系。并提出了一种很好的校正理论。经过实验检验,它能在图象列信息相同时,不管背景图象如何,均能使显示屏保持相同的亮度,使显示的视频图象的质量得以提高.     

XOR-based Visual Cryptography Schemes

A recent publication introduced a Visual Crypto (VC) system, based on the polarisation of light. This VC system has goodresolution, contrast and colour properties.Mathematically, the VC system is described by the XOR operation (modulo two addition). In this paper we investigate Threshold Visual Secret Sharing schemes associated to XOR-based VC systems. Firstly, we show that n out of n schemes with optimal resolution and contrast exist, and that (2,n) schemes are equivalent to binary codes. It turns out that these schemes have much better resolution than their OR-based counterparts. Secondly, we provide two explicit constructions for general k out of n schemes. Finally, we derive bounds on the contrast and resolution of XOR-based schemes. It follows from these bounds that for k<n, the contrast is strictly smaller than one. Moreover, the bounds imply that XOR-based k out of n schemes for even k are fundamentally different from those for odd k.AMS Classification: 94A60     

Molecular mechanisms of visual excitation: A concatenation of nonlinear cellular processes

The mechanism of transduction in the photoreceptor remains the least understood link in the chain of molecular and cellular processes that are responsible for vision. This paper reviews the molecular species and processes involved in the absorption of light and subsequent biochemical responses of the photoreceptor cell. The cells' electrical response and the probable ionic basis of the response are described. We develop the outlines of a mechanistic model that attempts to link biochemical and electrophysiological responses of photoreceptors of the vertebrate retina.     

基于人眼光学质量的球差和离焦补偿关系研究

基于人眼光学质量客观评估方法,通过实测的屈光手术前后人眼波前像差数据,研究和分析了球差与离焦的补偿关系。结果表明,对于屈光手术前的近视眼患者,离焦与球差的补偿关系并不明显。对于屈光手术后的近视眼患者,离焦与球差的补偿关系明显,残余离焦可以部分补偿由于高阶像差增加引起的视功能降低。当残余离焦处于0.27D～0.74D范围内,离焦与高阶像差组合后,光学质量提高了20%以上。在屈光手术方案设计时,预留0.27D～0.74D的离焦时,可以使80%的人眼达到更好的视觉矫正效果。     

基于可见-近红外光谱技术与BP-ANN算法的污水类型鉴定

提出了一种基于可见-近红外光谱技术与BP人工神经网络(BP-ANN)算法快速进行污水类型鉴定的新方法.以FieldSpec(R)3地物光谱仪采集了4种污水样品的光谱数据,共168份,随机将其分成校正集(132份)和检验集(36份).分别采取全波段(400～2450 nm)与择取波段(400～1800 nm)两种方法建立模型进行分析.光谱经S.Golay平滑和标准归一化(SNV)处理后,以主成分分析法(PCA)降维.将降维所得的前9个主成分数据作为BP-ANN的输入变量,污水类型作为输出变量,建立3层BP-ANN鉴别模型.利用36个未知样对模型进行检验.结果表明:两类模型预测准确率均高达100%,且择取波段模型比全波段模型具有更高的预测精度.说明利用可见-近红外技术结合BP-ANN算法进行污水类型的快速、无污染鉴定是可行的,且波段筛选是优化模型的有效方法之一.     

基于模糊积分的可信度融合目标检测

针对可见光和红外热图像序列中的远距离目标检测问题,提出了一种基于模糊积分的可信度融合的目标检测方法。该方法通过帧间差累积,在两种传感器的图像中确定了运动目标区域,根据目标区域的强度分析,定义了目标可信度度量,进而利用模糊积分融合函数进行可信度融合以实现目标检测。试验结果表明模糊积分分类效果较好,从而证明了该方法在目标识别中的可靠性和可信度,具有一定的实用性。