基于双向非局部模型的图像插值

如何充分利用自然图像的自相似特性来进行图像恢复是近年来图像处理领域研究的热点问题之一.事实上,自相似体现为两方向相似结构,当一组相似块被安排成矩阵时,该矩阵的行与行、列与列之间均存在相似性.为此,提出一种双向非局部模型,该模型对称地利用了两方向相似结构,此外,该模型直接将相似块作为字典来表示每一图像块且对表示系数进行Tikhonov正则化.将双向非局部模型用于图像插值,获得了目前最好的恢复效果,无论峰值信噪比还是视觉质量,相比已有算法都有很大地提高.     

基于HVS与PCNN的彩色图像增强

提出一种新的彩色图像增强方法:将RGB空间转化为HIS空间后保持H分量不变,对S分量进行指数拉伸;而对I分量则利用脉冲耦合神经网络的特性将原始图像的灰度信息与空间信息一起耦合到神经元的内部活动项,然后利用人类视觉系统模型构建动态阈值与其比较得到增强图像.实验证实了这种方法的有效性.     

基于反馈机制的抗混合噪声的盲图像源分离方案

针对基于稀疏成分分析的盲图像源分离算法无法解决混合噪声问题,本文提出一种采用反馈机制的盲源分离算法.通过小波域稀疏成分分析和置零反馈的方法,逐次分离出各支路信号.实验结果表明,该方法无需大量的迭代运算,与传统稀疏成分分析法相比,能有效地分离高斯白噪声参与的混合图像,与经典快速独立成分分析法相比,取得了更高的分离精度.     

求同存异的非局部图像去噪

非局部方法是近年来图像恢复领域最重要的方法之一,这类方法的关键在于如何理解和描述自然图像的自相似性质.本文通过对自相似性质的深入分析,提出描述自相似性质的两个原则：1）＂两方向＂原则：利用隐含在图像中的两方向相关结构.2）＂求同存异＂原则：利用相似性的同时保持各相似块之间的相对差异.然后利用奇异值分解建立了一种符合这两个原则的非局部图像去噪模型.该模型能获得很好的去噪效果.     

基于改进K均值的脸形聚类算法

针对海量人脸图像数据库检索时顺序匹配速度慢等问题,提出把聚类技术应用于数据库预分类,利用脸形特征对人脸图像自动聚类.首先用改进主动形状模型提取脸形特征,再用改进K-均值算法对人脸图像进行聚类.使用Hausdorff距离计算两个特征点集的相似度.实验表明,该算法的聚类结果比较稳定、精确且符合人类视觉认知特性.     

基于卷积神经网络与Bayesian决策的图像识别与分类记忆建模

当前热门图像分类方法大多侧重在分类能力,忽视识别新事物,然而人类认识事物时侧重认识,只在细小之处重视分类,这一点与人类记忆机制密切相关.尽管目前有许多记忆建模理论被相继提出,但大多以单词列表的形式学习,对自然图像列表的研究有限.基于此,本文提出了基于卷积神经网络与Bayesian决策的图像识别分类记忆建模方法,首先利用卷积神经网络提取图像特征,并采用二进制形式存储特征向量;然后进行视觉图像的表达,存储与提取记忆建模,将测试图像特征向量与所有已存储特征向量进行匹配对比,计算似然率值;最后在所有似然率基础上计算测试图像是新类别的几率,若该几率大于某个阈值则判别其为新类别;反之,利用Bayesian决策规则进行.图像分类.在Caltech-101与Caltech-256数据库上的实验表明所提方法能很好地应用于图像识别分类任务中.其击中率比目前代表性的稀疏表达分类(SRC)以及极限学习机(ELM)方法高,且虚报率比其他两种方法低的多.     

基于PCNN捕获-抑制机制的图像分类器设计

针对医学图像识别的重要性,本文提出了一种结合彩色图像特征场和基于脉冲耦合神经网络捕获-抑制机制的图像分类器设计方法.首先,构造符合人类视觉的彩色图像特征场;然后,利用脉冲耦合神经网络设计分类器实现对图像的识别.最后,应用于JAFFE人脸表情数据库和中医舌象的分类识别中,并与ADABOOST方法和SVM方法进行比较.实验结果表明,本文方法可以获得更好的医学图像识别效果.     

视差计算的层级模型

生物的立体感知是由一个层级网络完成的:从初级视皮层到高层区域,神经元感受野逐步增大,局部立体感知逐步变为全局立体感知.视皮层中存在大量对视差敏感的神经元,其中V1区单个神经元的视差选择特性可用视差能量模型来描述.文中从生物的立体感知过程出发,提出了一种计算图像视差的层级模型,主要贡献有:(1)提出了一种符合心理学实验结果的归一化视差能量模型,减弱了图像对比度变化对神经元视差响应能量的影响;(2)利用视皮层视差功能柱的性质,提出了一种不同倾向视差神经元的汇聚方法;(3)根据不同脑皮层之间的连接关系,提出了一种两层网络结构来解决V1区神经元编码视差的歧义问题.文中方法可以有效提高纹理重复和纹理不丰富区域的视差计算精度.     

基于特征识别的增强现实跟踪定位算法

文中提出一种在场景自然特征识别基础上采用关键帧匹配的增强现实跟踪注册算法,实现了基于计算机视觉的户外大场景范围内的实时跟踪定位.算法针对关键帧匹配中的宽基线特征匹配问题,提出采用随机树的模式分类方法实现场景自然特征的离线训练和在线实时精确匹配.同时采用Kalman滤波器对参数估计结果进行平滑解决视觉跟踪定位中的抖动问题.以该算法为核心构建出基于视频透视式头盔显示器的移动增强现实系统,并将其成功应用于圆明园大水法景观的数字重现.真实场景实验验证表明,该方法具有实时、鲁棒的优点,适用于户外跟踪定位.     

一种基于改进 Surendra背景更新算法的运动目标检测

建立背景模型时,采用动态阈值的方法,以减少噪声和光照强度变化等外界环境的影响;在运动目标检测中,针对传统方法检测到的运动目标偏大的问题,提出一种改进的三帧差分检测方法.在形态学处理过程中,采用自适应方法选择结构元素的大小,对二值图像进行去噪处理,以得到平滑图像.实验证明,改进后的方法与传统背景模型建立方法相比,能更准确地检测到运动目标区域,并实时更新背景模型     

Effect of adapting target size on the gain of the surround response mechanism in X- and Y-cells in cat retina.

The adaptation field of the surround mechanism of X and Y retinal ganglion cells in the cat was assessed with variable size, unmodulated adapting spots. Both an on-inhibition measure and an off-discharge measure of surround gain was used. Results suggest that the surround mechanism in Y-cells is strongest in the receptive field middle but weak or nonexistent in the middle of X-cell receptive fields.     

Effect of adapting target size on the gain of the surround response mechanism in X- and Y-cells in cat retina

Summary The adaptation field of the surround mechanism of X and Y retinal ganglion cells in the cat was assessed with variable size, unmodulated adapting spots. Both an on-inhibition measure and an off-discharge measure of surround gain was used. Results suggest that the surround mechanism in Y-cells is strongest in the receptive field middle but weak or nonexistent in the middle of X-cell receptive fields.Acknowledgment. This research was supported by Public Health Service grant No. EY 00701.     

The vomeronasal system

In addition to the main olfactory system, many vertebrates possess a vomeronasal system that conveys more specialized chemosensory information. Unlike the airborne, volatile stimuli detected by the main olfactory system, vomeronasal stimuli are typically proteins of the lipocalin family which bind small, volatile ligands. Despite the smaller number of vomeronasal receptor types, the projection patterns of the vomeronasal receptor neurons to multiple glomeruli in the accessory olfactory bulb appear to be more complicated than those of the main olfactory system. The vomeronasal system has a direct sub-neocortical projection to hypothalamic areas that mediates specific behavioural and hormonal responses to pheromonal stimuli. However, the integration and transmission of this information can be modulated by learning mechanisms. The aim of this article is to outline some of the functions of the vomeronasal system, and in particular to comment on recent advances in our understanding of how vomeronasal information is coded and processed.     

Implication of GABA in the spatial organization of the receptive fields of retinal ganglia cells of frogs]

An histoautoradiographic study of the Chicken and Frog retinae was made following an intravitreal injection if tritiated GABA. This amino acid was especially fixed at the amacrine and horizontal cell level. It has been suggested (Werblin, 1974) that horizontal cells might be responsible for the spatial organization of the receptive fields of retinal ganglion cells. GABA could thus be involved in bringing about of the centre-surround arrangement of the latter receptive fields. We have shown that picrotoxin, a specific GABA-antagonist, does indeed provoke marked changes in the activity of retinal ganglion cells. It provokes a decrease in the response threshold, a facilitation of spontaneous activities, and a marked increase in the receptive field area of the ON-OFF ganglion cells: from 12 at the start, it reaches 24 under the effect of picrotoxin.     

POLSAR图像杂波L分布建模及其参数估计

基于乘积模型的杂波统计建模是进行高分辨极化合成孔径雷达(POLSAR)图像非均匀区域杂波统计特性分析的有效方法,其核心在于纹理分量随机分布的类型选择.广义伽马分布(GΓD)是一种普适性很强的分布类型,Weibull分布、伽马分布、逆伽马分布等都是GΓD分布的特例,因此基于GΓD分布的图像乘性杂波建模既满足杂波建模的高精度要求,同时成为各种雷达杂波辨识的有效工具.本文首先推导了服从GΓD分布的随机变量的高阶矩特征及其对数累积量(MLC),利用纹理分量服从GΓD分布情形构建了乘积相干斑模型,得到了适用于POLSAR图像处理的L分布杂波多视处理的协方差矩阵的概率分布函数,同时推导了其高阶矩特征及其对数累积量,提出了基于对数累积量的L分布参数估计新方法.针对样本数较少的情况下对数累积量参数估计失效问题提出了基于混合矩(MME)的参数估计方法来解决.然后给出了不同分布的高阶矩和对数累积量,通过二三阶对数累积量关系图辨析了常用分布与L分布的内在关系,得到了L分布是目前乘积模型中适用范围较为广泛的统计分布的结论.最后用仿真数据验证了理论推导的正确性,并将基于对数累积量的参数估计方法与已有方法进行了比较,结果证明新参数估计方法具有更高的估计精度和运算效率;另外,还用实测数据进行了统计模型检验,其结果验证了理论推导的正确性.极化SAR中多视图像L分布杂波的统计建模及其参数估计方法为极化SAR目标检测和识别等领域的新技术研究提供了新手段.     

Ras proteins: paradigms for compartmentalised and isoform-specific signalling

Ras GTPases mediate a wide variety of cellular processes by converting a multitude of extracellular stimuli into specific biological responses including proliferation, differentiation and survival. In mammalian cells, three ras genes encode four Ras isoforms (H-Ras, K-Ras4A, K-Ras4B and N-Ras) that are highly homologous but functionally distinct. Differences between the isoforms, including their post-translational modifications and intracellular sorting, mean that Ras has emerged as an important model system of compartmentalised signalling and membrane biology. Ras isoforms in different subcellular locations are proposed to recruit distinct upstream and downstream accessory proteins and activate multiple signalling pathways. Here, we summarise data relating to isoform-specific signalling, its role in disease and the mechanisms promoting compartmentalised signalling. Further understanding of this field will reveal the role of Ras signalling in development, cellular homeostasis and cancer and may suggest new therapeutic approaches.     

Sensory basis of bird orientation

Sensory information which may be essential for the complex process of orientation of birds is described in this article. The use of vibrational, visual, chemical, olfactory, magnetic cues and their receptive mechanisms, as far as they are known, are explained. Special reference is given to the behavioral and physiological aspects of magnetic sensitivity.     

Sensory basis of bird orientation

Summary Sensory information which may be essential for the complex process of orientation of birds is described in this article. The use of vibrational, visual, chemical, olfactory, magnetic cues and their receptive mechanisms, as far as they are known, are explained. Special reference is given to the behavioral and physiological aspects of magnetic sensitivity.