感受野视觉特征整合模型及应用

 采用基本ICA模拟视觉感知机制对自然图像分解得到的图像基函数在空间排列上是混乱的,这与视觉生理机制相互矛盾.模拟视皮层感受野间的信息整合机制,建立了新的计算模型.针对基于内容的图像故障区域检测问题,提出了相应的高效率少样本检测算法. 首先,以列车正常和故障图像序列作为训练数据,利用拓扑ICA方法学习图像基函数,由此得到的独立分量系数作为神经元响应,然后模拟同步振荡机制选择响应强烈的神经元,输出其对应的内容,最后通过自动对比实现图像故障区域的快速定位.实验结果表明,与传统方法相比较,引入视觉信息整合机制的新模型及其算法能够提高故障检测率.     

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.     

Melanopsin-expressing ganglion cells in primate retina signal colour and irradiance and project to the LGN

Human vision starts with the activation of rod photoreceptors in dim light and short (S)-, medium (M)-, and long (L)- wavelength-sensitive cone photoreceptors in daylight. Recently a parallel, non-rod, non-cone photoreceptive pathway, arising from a population of retinal ganglion cells, was discovered in nocturnal rodents. These ganglion cells express the putative photopigment melanopsin and by signalling gross changes in light intensity serve the subconscious, 'non-image-forming' functions of circadian photoentrainment and pupil constriction. Here we show an anatomically distinct population of 'giant', melanopsin-expressing ganglion cells in the primate retina that, in addition to being intrinsically photosensitive, are strongly activated by rods and cones, and display a rare, S-Off, (L + M)-On type of colour-opponent receptive field. The intrinsic, rod and (L + M) cone-derived light responses combine in these giant cells to signal irradiance over the full dynamic range of human vision. In accordance with cone-based colour opponency, the giant cells project to the lateral geniculate nucleus, the thalamic relay to primary visual cortex. Thus, in the diurnal trichromatic primate, 'non-image-forming' and conventional 'image-forming' retinal pathways are merged, and the melanopsin-based signal might contribute to conscious visual perception.     

Molecular identification of a retinal cell type that responds to upward motion

The retina contains complex circuits of neurons that extract salient information from visual inputs. Signals from photoreceptors are processed by retinal interneurons, integrated by retinal ganglion cells (RGCs) and sent to the brain by RGC axons. Distinct types of RGC respond to different visual features, such as increases or decreases in light intensity (ON and OFF cells, respectively), colour or moving objects. Thus, RGCs comprise a set of parallel pathways from the eye to the brain. The identification of molecular markers for RGC subsets will facilitate attempts to correlate their structure with their function, assess their synaptic inputs and targets, and study their diversification. Here we show, by means of a transgenic marking method, that junctional adhesion molecule B (JAM-B) marks a previously unrecognized class of OFF RGCs in mice. These cells have asymmetric dendritic arbors aligned in a dorsal-to-ventral direction across the retina. Their receptive fields are also asymmetric and respond selectively to stimuli moving in a soma-to-dendrite direction; because the lens reverses the image of the world on the retina, these cells detect upward motion in the visual field. Thus, JAM-B identifies a unique population of RGCs in which structure corresponds remarkably to function.     

一种基于内容的图像比率可分级编码方法的研究

多分辨率环境下 的多媒体应用导致了基于内容的视频和图像编码的需求。提出一种基于内容的图像比率可分级性编码方法，方法具有以下特点：以不同的任意形状的ROI（Regions of Interest)分别进行独立的 小波分散和编码；采用逐次逼近的量化模式，以形成潜入式码流；解码时，用户可选择相应的ROI进行比率可分级解码。它在一定程度上满足了JPEG－2000中的ROI编码和MPEG－4中的静态图像纹理编码的需求。试验结果验证了方法的有效性。     

Receptive field dynamics in adult primary visual cortex.

The adult brain has a remarkable ability to adjust to changes in sensory input. Removal of afferent input to the somatosensory, auditory, motor or visual cortex results in a marked change of cortical topography. Changes in sensory activity can, over a period of months, alter receptive field size and cortical topography. Here we remove visual input by focal binocular retinal lesions and record from the same cortical sites before and within minutes after making the lesion and find immediate striking increases in receptive field size for cortical cells with receptive fields near the edge of the retinal scotoma. After a few months even the cortical areas that were initially silenced by the lesion recover visual activity, representing retinotopic loci surrounding the lesion. At the level of the lateral geniculate nucleus, which provides the visual input to the striate cortex, a large silent region remains. Furthermore, anatomical studies show that the spread of geniculocortical afferents is insufficient to account for the cortical recovery. The results indicate that the topographic reorganization within the cortex was largely due to synaptic changes intrinsic to the cortex, perhaps through the plexus of long-range horizontal connections.     

基于字典统计耦合归一化多重距离的图像检索算法

针对当前彩色图像检索技术容易受到色彩干扰,鲁棒性不强等不足,提出了一种字典统计耦合归一化多重距离的彩色图像检索算法.首先,将图像量化并转换成一维信号;然后,引入字典统计,将一维信号进行字典编码,并计算编码后的图像多样值;在归一化字典距离的基础上,嵌入字典编码图像的多样值,从而定义了归一化多重距离(NMD)的相似度量准则,利用NMD对查询图像与数据库图像的多样值进行比较与识别,搜索出与查询图像具有相同特征的最相似图像,完成目标检索.在COREL数据库的实验结果表明:相对于当前常用的检索技术,该检索算法具有更高的查准率和查全率,可对彩色图像完成精确检索,有效减低了色彩对检索性能的干扰.     

采用预测编码的图像隐写分析算法

通过分析图像无损压缩过程中预测编码的原理,提出了一种基于预测编码方法的空域图像隐写分析算法。该算法的基本思想是利用空域隐写会破坏图像相邻像素之间的相关性,而预测编码是利用图像相邻像素之间的相关性来预测的,这样隐写后图像的预测误差将会大于载体图像的预测误差,通过计算隐写前后图像的预测误差来分析图像中是否含有秘密信息。实验表明:基于预测编码的图像隐写分析具有较好的隐写分析能力。     

采用HEVC的精细可分级编码

针对新一代的视频编码标准HEVC(high efficient video coding),提出一种改进的精细可分级编码方案.该方案的基本层采用HEVC的编码器,提高了基本层的编码效率;通过统计编码单元的分割方式,自适应找到图像中的细节区域;采用选择增强技术,提高细节区域的图像质量.实验结果表明:编码方案能够精细匹配信道带宽的变化,且基本层与增强层相互独立,不会带来误差传递;利用HEVC编码单元的分割方式,可以自适应找到视频序列图像中的细节区域,对运动区域进行提升,视频图像的主观质量有了很大的改善.     

Functions of the colour-opponent and broad-band channels of the visual system

The colour-opponent and broad-band channels of the primate visual system originate in the retina and remain segregated through several neural stations in the visual system. Until now inferences about their function in vision have been based primarily on studies examining single-cell receptive field properties which have shown that the colour-opponent retinal ganglion cells have small receptive fields, produce sustained responses and receive spatially segregated inputs from different cone types; the broad-band cells have large receptive fields, respond transiently and receive cone inputs that are not spatially separated. We have now examined the visual capacities of rhesus monkeys before and after interrupting either of these channels with small lesions at the lateral geniculate nucleus. Here we report that the colour-opponent channel is essential for the processing of colour, texture, fine pattern and fine stereopsis, whereas the broad-band channel is crucial for the perception of fast flicker and motion. Little or no deficits were found in brightness and coarse-shape discrimination, low spatial frequency stereopsis and contrast sensitivity after the disruption of either of the channels.     

基于方向剖分的小波域分形图象压缩

该文提出一种新的分形和小波混合编码方法－基于方向剖分的小波域分形图象压缩方法：在将图象分解到小波域后，对各子图象，根据其所占能量的大小和所代表的方向纹理信息，采用不同大小、形状与类别的定义域块和值域块进行分形编码，经实验证明，这种基于方向剖分的小波域分形编码方法，在保证一定的还原图象质量的前提下，可以大大地提高了分形编码的速度，如果结合多层小波分解的分形压缩编码的预测法和零树法，还能进一步提高压缩率，从而使分形图象编码向实时应用迈进了一步。     

视网膜神经节细胞非经典感受野及其方位倾向性的模型研究

建立了包含无长突细胞相互抑制网络的视网膜神经节细胞的三层网络模型,以验证李朝义等提出的非经典感受野可能源于无长突细胞相互抑制的假说,并探讨神经节细胞感受野的方位倾向性的可能机制．模拟结果表明,通过无长突细胞之间的相互抑制可以形成神经节细胞的非经典感受野．模型模拟了神经节细胞的中心区和大周边区的方位倾向性及其相互作用,结果提示神经节细胞的方位倾向性可能主要源于神经节细胞树突野的空间分布．     

基于HMAX模型和非经典感受野抑制的轮廓提取

针对图像里处于复杂纹理背景中物体的轮廓提取正确率低的问题,首先研究了基于非经典感受野抑制的轮廓提取算法和HMAX模型,然后利用HMAX模型所具备的具有基本视皮层功能结构的优点,弥补了前者所依据的生物学视觉结构比较简单的不足,最后提出并实现了基于HMAX模型和非经典感受野抑制的轮廓提取算法。通过与Canny算子和非经典感受野抑制的轮廓提取算法的评估比较,表明本文算法有效提高了轮廓提取的正确率。     

一种用于高动态图像的自适应编码方法的改进研究

高动态图像亮度变化范围较大,当前编码方法采用DCT技术描述高动态图像信号的变换,无法准确描述图像细节、频率突变的高动态图像信号,导致编码后图像的极亮和极暗区域可视性低、细节不丰富。提出一种新的用于高动态图像的自适应编码方法,分析了高动态图像编码过程,通过小波变换实现高动态图像信号描述形式的自适应变换,对高动态图像进行多尺度分解,获取不同尺度的高动态图像信息,得到子带图像;对各子带图像的小波系数进行量化,以达到提高图像码率的目的。在此基础上,选用Log Luv(TIFF)编码方法,对高动态图像信息数据间的统计相关性进行滤除,完成高动态图像的自适应编码。实验结果表明,采用所提方法对高动态图像进行编码后,得到的图像质量较高;且在编码效率较低的情况下,仍能保持较高的图像质量,编码性能明显优于预测编码方法。     

VTC-KSVD:一种融合视觉特征与标签一致性的多标签图像标注方法

提出一种融合视觉特征及标签一致性的多标签图像标注方法VTC-KSVD.首先通过K均值奇异值分解（KSVD）法建立图像的标签一致性模型TC-KSVD,然后将多视图特征融合在该模型中.该方法既利用了训练样本的类标与编码系数的判别式模型,又利用了训练样本的标签与编码系数的关系,增加了字典的判别性,提高了标注性能.在Corel5K数据集上的实验结果表明,融合了多视图视觉特征与标签一致性的VTC-KSVD方法可以较为准确地找到视觉特征与语义特征均相似的图像近邻,能明显提升多标签图像的标注性能,并能有效缓解训练数据有限而引起的稀疏性问题.      

基于局部变化率的无损图像压缩方法

介绍一种实现图像无损压缩的预测编码算法。该算法在预测阶段利用像素值的局部变化率来对预测模型进行自适应调整,在编码阶段采用误差反馈技术进一步降低误差图像的信息熵。对标准图像的仿真测试结果表明,该算法的性能明显优于国际标准化组织JPEG－LS提供的标准无损压缩算法。     

Dynamic receptive field plasticity in rat spinal cord dorsal horn following C-primary afferent input

The central terminals of cutaneous primary afferent neurons are spatially ordered in the dorsal horn in a highly organized fashion such that a point-to-point map represents the body surface. This afferent terminal somatotopic map correlates with the map of the receptive fields of the cells on which they terminate. The location, size and modality of the cutaneous receptive fields of dorsal horn neurons necessarily depend upon the anatomical presence of afferent nerve fibres which deliver information from the periphery, directly or indirectly, to the cells. However the receptive field size and modality of a cell do not depend only on anatomical connections. Excitatory and inhibitory interneurons, descending influences and facilitations or depressions of synaptic contacts can alter receptive field properties. Here we show that prolonged and substantial cutaneous receptive field changes can be produced by brief inputs from peripheral unmyelinated afferent fibres.     

Anticipation of moving stimuli by the retina

A flash of light evokes neural activity in the brain with a delay of 30-100 milliseconds, much of which is due to the slow process of visual transduction in photoreceptors. A moving object can cover a considerable distance in this time, and should therefore be seen noticeably behind its actual location. As this conflicts with everyday experience, it has been suggested that the visual cortex uses the delayed visual data from the eye to extrapolate the trajectory of a moving object, so that it is perceived at its actual location. Here we report that such anticipation of moving stimuli begins in the retina. A moving bar elicits a moving wave of spiking activity in the population of retinal ganglion cells. Rather than lagging behind the visual image, the population activity travels near the leading edge of the moving bar. This response is observed over a wide range of speeds and apparently compensates for the visual response latency. We show how this anticipation follows from known mechanisms of retinal processing.     

连续性健美操动作图像视觉误差校正方法分析

为了解决传统方法校正后图像容易出现模糊或锯齿效应,视觉误差校正效果不好的问题,研究了一种新的连续性健美操动作图像视觉误差校正方法。通过JAI相机(丹麦)逐行扫描电荷耦合器件(CCD)相机转换成视频信号,得到的信号通过图像采集卡采集,转化成PC机可处理的数字信号,通过全息投影获取连续性健美操动作图像。按照视觉误差完成对图像像素集视差函数的配准,通过向量量化技术实现对图像的亚像素级模板匹配,对经处理的连续性健美操动作图像进行分形编码处理,为视觉误差校正提供依据。对连续性健美操动作图像进行编码和解码处理,求出和原始图像间的误差,获取误差图像,对其进行插值处理,获取误差补偿结果,实现连续性健美操动作图像视觉误差校正。结果表明:经所提方法处理后的图像无显著的视觉误差,峰值信噪比和结构相似性很高。可见所提方法主观和客观评价结果好。