A computational coding model for saliency detection in primary visual cortex

This study researches the coding model adaptive for information processing of the bottom-up attention mechanism.We constructed a coding model satisfying the neurobiological constraints of the primary visual cortex.By quantitatively changing the coding constraints,we carried out experiments on images used in cognitive psychology and natural image sets to compare the effects on the saliency detection performance.The experimental results statistically demonstrated that the encoding of invariant features and representation of overcomplete bases is advantageous to the bottom-up attention mechanism.     

Dynamic properties of neurons in cortical area MT in alert and anaesthetized macaque monkeys.

In order to see the world with high spatial acuity, an animal must sample the visual image with many detectors that restrict their analyses to extremely small regions of space. The visual cortex must then integrate the information from these localized receptive fields to obtain a more global picture of the surrounding environment. We studied this process in single neurons within the middle temporal visual area (MT) of macaques using stimuli that produced conflicting local and global information about stimulus motion. Neuronal responses in alert animals initially reflected predominantly the ambiguous local motion features, but gradually converged to an unambiguous global representation. When the same animals were anaesthetized, the integration of local motion signals was markedly impaired even though neuronal responses remained vigorous and directional tuning characteristics were intact. Our results suggest that anaesthesia preferentially affects the visual processing responsible for integrating local signals into a global visual representation.     

Neural mechanisms of unconscious visual motion priming

The neural correlates of the motion priming were examined in normal young subjects using event-related brain potentials (ERPs) and functional magnetic resonance imaging (fMRI). Visual motion perception can be uncon-sciously biased in favor of a particular direction by a pre-ceding motion in that direction. Motion priming first in-volved an enhancement of ERP amplitude about 100 ms fol-lowing the onset of motion. The amplitudes of ERP compo-nents after 350 ms were also increased. The fMRI results suggest that the early-latency effect reflects modulation of neural responses in extrastriate cortex. Higher-level visual processing areas, including cortical regions MT/MST and the intraparietal cortices were also activated. The findings provide direct evidence that unconscious priming of motion perception is the result of interaction of direction-selective neural responses to motion stimuli. The results cannot be accounted for by refractoriness of neural responses, but in-stead support a theory of motion priming based on motion opponency, as proposed in computational models.     

基于图像矩的机器人视觉伺服

区别于图像的简单几何特征,利用图像的全局特征描述子－图像矩特征作为图像作征信息,推导了矩特征变化量与相对位姿变化量之间的关系矩阵,即图像雅可比矩阵。针对平面视觉伺服控制,利用了所推导的基于矩特征的图像雅可比矩阵,在不需要知道目标的成像高度、摄像机焦距及不需要精确标定的摄像机外部参数的情况下实现了基于图像的视觉伺服控制,最后给出了仿真与实验例子,表明了矩特征的有效     

Imaging cortical correlates of illusion in early visual cortex

Exploring visual illusions reveals fundamental principles of cortical processing. Illusory motion perception of non-moving stimuli was described almost a century ago by Gestalt psychologists. However, the underlying neuronal mechanisms remain unknown. To explore cortical mechanisms underlying the 'line-motion' illusion, we used real-time optical imaging, which is highly sensitive to subthreshold activity. We examined, in the visual cortex of the anaesthetized cat, responses to five stimuli: a stationary small square and a long bar; a moving square; a drawn-out bar; and the well-known line-motion illusion, a stationary square briefly preceding a long stationary bar presentation. Whereas flashing the bar alone evoked the expected localized, short latency and high amplitude activity patterns, presenting a square 60-100 ms before a bar induced the dynamic activity patterns resembling that of fast movement. The preceding square, even though physically non-moving, created gradually propagating subthreshold cortical activity that must contribute to illusory motion, because it was indistinguishable from cortical representations of real motion in this area. These findings demonstrate the effect of spatio-temporal patterns of subthreshold synaptic potentials on cortical processing and the shaping of perception.     

Visualizing columnar architectures using high-field fMRI

Among the many neuroimaging tools available for studying human brain functions, functional magnetic resonance imaging (fMRI) is the most widely used today. One advantage of fMRI over other imaging techniques is its relatively high spatial resolution. High-resolution fMRI, with its superb signal-to-noise ratio and improved tissue-vessel specificity, has strengthened the capability of fMRI and allowed mapping of fine cortical architectures in the human brain. In this review, I will first explain the factors limiting the spatial specificity of the blood oxygenation level-dependent (BOLD) effect, based on which most of fMRI experiments are conducted, and the measures dealing with these factors, and then briefly introduce several high-resolution (sub-millimeter) studies on the functional organization of human primary visual cortex (V1), including mapping of ocular dominance columns, mapping of temporal frequency dependent domains and direct demonstration of tuning to stimulus orientation.     

Visualizing columnar architectures using high-field fMRI

Among the many neuroimaging tools available for studying human brain functions, functional magnetic resonance imaging (fMRI) is the most widely used today. One advantage of fMRI over other imaging techniques is its relatively high spatial resolution. High-resolution fMRI, with its superb signal-to-noise ratio and improved tissue-vessel specificity, has strengthened the capability of fMRI and allowed mapping of fine cortical architectures in the human brain. In this review, I will first explain the factors limiting the spatial specificity of the blood oxygenation level-dependent (BOLD) effect, based on which most of fMRI experiments are conducted, and the measures dealing with these factors, and then briefly introduce several high-resolution (sub-millimeter) studies on the functional organization of human primary visual cortex (V1), including mapping of ocular dominance columns, mapping of temporal frequency dependent domains and direct demonstration of tuning to stimulus orientation.     

一种用于图像检索的幂归一化深度卷积特征加权聚合方法

针对图像检索中基于部位的加权聚合(PWA)方法存在的视觉突发问题,提出一种幂归一化的深度卷积特征加权聚合方法。首先简化了原PWA方法中用于确定空间权重的归一化和幂变换操作,直接将所选择的有区分性的通道特征图作为空间权重矩阵,然后引入新的幂变换函数并选取合适的参数对加权聚合后的通道响应进行归一化处理,最后通过PCA降维和白化处理形成图像的全局特征表示形式。在4个标准数据库上的图像检索实验结果表明,该方法能有效调节PWA聚合特征响应的突发度并提高图像检索的准确率。     

错视在广告中的运用

错视是“对客观事物不正确的视知觉”,属于错觉的一种。在现实生活中,这种错视现象会经常发生,使得现实事物在我们的视觉中发生某种程度的变化,有时可得到美化,有时却产生变形丑化的效果。而对于以推销商品为目的的广告来说,恰当地运用视错觉,以提升产品的形象,或者强化产品在受众心里所形成的视觉效果,是非常重要的。如果运用得巧妙而且得当,往往会收到意想不到的效果。     

视觉神经表征中底-顶和顶-底的双向激活——来自人与猴的实验证据

长期以来,建立在坚实生理学证据之上的底-顶加工说和特征检测理论在视觉研究中占主导地位。对于顶-底加工人们只能靠一般常识,即知识或经验通过激活记忆中的神经表征影响视觉过程。但是近年来,来自人和猴的研究为顶-底的加工提供了实验证据。 首先位于猴腹侧加工系统内的物体和面孔视觉记忆表征,提供了神经编码是怎样创立、组织和再激活的最佳实验证据。联想性编码是通过学习由一些具有特殊功能的神经元建立的,这些神经元具有将时间性关联刺激的表征联系起来的能力。其次,不仅来自视网膜的底-顶信号,而且来自前额叶的顶-底信号都能触发联想性编码的提取,既可以作为有意识回忆的神经基础,又是顶-底加工影响视觉过程的基础。脑损伤病人研究、具有高时间分辨率的人类功能性核磁共振成像(functional magnetic resonance imaging, fMRI)和猴fMRI研究以及猴细胞电生理分析相结合,将进一步加强人们对视觉脑机制的全面理解。     

Feature-based attention influences motion processing gain in macaque visual cortex.

Changes in neural responses based on spatial attention have been demonstrated in many areas of visual cortex, indicating that the neural correlate of attention is an enhanced response to stimuli at an attended location and reduced responses to stimuli elsewhere. Here we demonstrate non-spatial, feature-based attentional modulation of visual motion processing, and show that attention increases the gain of direction-selective neurons in visual cortical area MT without narrowing the direction-tuning curves. These findings place important constraints on the neural mechanisms of attention and we propose to unify the effects of spatial location, direction of motion and other features of the attended stimuli in a 'feature similarity gain model' of attention.     

Perceptual consequences of centre-surround antagonism in visual motion processing

Centre-surround receptive field organization is a ubiquitous property in mammalian visual systems, presumably tailored for extracting image features that are differentially distributed over space. In visual motion, this is evident as antagonistic interactions between centre and surround regions of the receptive fields of many direction-selective neurons in visual cortex. In a series of psychophysical experiments we make the counterintuitive observation that increasing the size of a high-contrast moving pattern renders its direction of motion more difficult to perceive and reduces its effectiveness as an adaptation stimulus. We propose that this is a perceptual correlate of centre-surround antagonism, possibly within a population of neurons in the middle temporal visual area. The spatial antagonism of motion signals observed at high contrast gives way to spatial summation as contrast decreases. Evidently, integration of motion signals over space depends crucially on the visibility of those signals, thereby allowing the visual system to register motion information efficiently and adaptively.     

基于多尺度特征融合的图片情感分布学习

视觉情感分析旨在分析人们对视觉刺激的情感反映,近年来受到了共享平台和网络社交等多媒体视觉数据相关领域的关注.传统的图片情感分析侧重于单标签的情感分类,忽略了图片表达的情感的复杂性和图像潜在的情绪分布信息,不能体现出图片所表达的不同情绪之间的相关性.针对以上问题,首先采用ViT和ResNet网络进行全局和局部融合的多尺度情感特征提取,通过主导情绪分类和标签分布学习进行图片情感识别,充分表征图片的复杂情感.在公开的Flickr_LDL数据集和Twitter_LDL数据集上取得了显著的效果,证明了提出方法的有效性.     

基于视觉感知的模糊C均值聚类算法

针对传统模糊C均值聚类算法对结构复杂图像分割效果不理想的问题,提出一种基于视觉感知的模糊C均值聚类算法.首先,在分析视皮层神经元感受野性质的基础上,建立视神经元细胞响应函数来计算图像的结构特征.其次,定义一种斜坡函数从仿生学的角度来模拟人眼对相对亮度变化的感知,用来计算图像中像素点与聚类中心点之间的差异.所提模型充分考虑了邻域刺激对中心神经元影响的方向性、位置相对性和周期性,比较精确地描述了图像的结构信息,有效地抑制了噪声和复杂纹理的干扰.实验结果表明,本文算法克服了传统模糊C均值聚类算法的缺点,实现了具有复杂背景图像的精确分割.     

Capacity limit of visual short-term memory in human posterior parietal cortex

At any instant, our visual system allows us to perceive a rich and detailed visual world. Yet our internal, explicit representation of this visual world is extremely sparse: we can only hold in mind a minute fraction of the visual scene. These mental representations are stored in visual short-term memory (VSTM). Even though VSTM is essential for the execution of a wide array of perceptual and cognitive functions, and is supported by an extensive network of brain regions, its storage capacity is severely limited. With the use of functional magnetic resonance imaging, we show here that this capacity limit is neurally reflected in one node of this network: activity in the posterior parietal cortex is tightly correlated with the limited amount of scene information that can be stored in VSTM. These results suggest that the posterior parietal cortex is a key neural locus of our impoverished mental representation of the visual world.     

Image Post-Processing Method for Visual Data Mining

Visual data mining is one of important approach of data mining techniques. Most of them are based on computer graphic techniques but few of them exploit image-processing techniques. This paper proposes an image processing method, named RNAM （resemble neighborhood averaging method）, to facilitate visual data mining, which is used to post-process the data mining result-image and help users to discover significant features and useful patterns effectively. The experiments show that the method is intuitive, easily-understanding and effectiveness. It provides a new approach for visual data mining.     

Identifying natural images from human brain activity

A challenging goal in neuroscience is to be able to read out, or decode, mental content from brain activity. Recent functional magnetic resonance imaging (fMRI) studies have decoded orientation, position and object category from activity in visual cortex. However, these studies typically used relatively simple stimuli (for example, gratings) or images drawn from fixed categories (for example, faces, houses), and decoding was based on previous measurements of brain activity evoked by those same stimuli or categories. To overcome these limitations, here we develop a decoding method based on quantitative receptive-field models that characterize the relationship between visual stimuli and fMRI activity in early visual areas. These models describe the tuning of individual voxels for space, orientation and spatial frequency, and are estimated directly from responses evoked by natural images. We show that these receptive-field models make it possible to identify, from a large set of completely novel natural images, which specific image was seen by an observer. Identification is not a mere consequence of the retinotopic organization of visual areas; simpler receptive-field models that describe only spatial tuning yield much poorer identification performance. Our results suggest that it may soon be possible to reconstruct a picture of a person's visual experience from measurements of brain activity alone.     

Mapping multiple features in the population response of visual cortex

Stimulus features such as edge orientation, motion direction and spatial frequency are thought to be encoded in the primary visual cortex by overlapping feature maps arranged so that the location of neurons activated by a particular combination of stimulus features can be predicted from the intersections of these maps. This view is based on the use of grating stimuli, which limit the range of stimulus combinations that can be examined. We used optical imaging of intrinsic signals in ferrets to assess patterns of population activity evoked by the motion of a texture (a field of iso-oriented bars). Here we show that the same neural population can be activated by multiple combinations of orientation, length, motion axis and speed. Rather than reflecting the intersection of multiple maps, our results indicate that population activity in primary visual cortex is better described as a single map of spatiotemporal energy.     

物联网下大区域校园智能视觉特征定位技术仿真

为了解决传统技术易受外界干扰,造成视觉特征存在缺失,影响定位结果,且仅可应用于颜色特征显著的视觉特征定位的弊端,通过SURF法和Euler距离匹配研究了一种物联网下大区域校园智能视觉特征定位技术。通过物联网技术对监控的大区域校园图像进行智能采集,给出物联网视觉传感器分布情况。针对采集图像进行预处理,增强图像干扰抑制能力。把图像当成包,把分割后图像块当成包中的示例,在大区域内为某视觉图像确定最优标注。在此基础上,通过SURF算法对视觉特征点进行检测,利用Euler距离实现物联网下大区域校院智能视觉特征匹配定位。结果表明：所提技术检测特征无显著差异,具有不变性;对白天校园人行道区域进行视觉特征定位,定位误差低;对夜间校园主干道区域进行视觉特征定位,定位误差较白天无显著差异。可见所提技术视觉特征定位精度高。     

基于图像矩的眼注视研究

利用图像的全局特征描述子－图像矩特征作为图像特征信息,从而构成视觉伺服,实现对动目标的注视跟踪,俯仰由末执行器完成,摇摆由腰的旋转完成,最后给出了实验结果,表明了矩特征的有效性以及五自由度机器人实现注视跟踪的可能性。