织物纹理的简单视神经细胞感受野的选择特性

研究了织物纹理的简单视神经细胞感受野的选择特性。分别选择具有明显方向性和周期性以及周期性和方向性不显著的两类代表织物纹理为研究对象,采用基于独立分量分析的视觉模型估计出简单视觉细胞的感受野。通过对感受野兴奋区域重心、方向角和面积的分析,以有效表示感受野在位置、方向和空间频率的选择特性。基于织物纹理与自然纹理两类纹理的简单细胞感受野选择性比较实验显示,相比于自然纹理,织物纹理的简单细胞感受野单元中的兴奋区域在方向、位置和空间频率上具有更加显著的选择特性。实验结果表明,提出的简单视觉细胞感受野选择特性研究方法在一定程度上揭示了初级视皮层对织物纹理图像处理的神经机制,从视觉感知层面反映了织物纹理的视觉特性。     

引入纹理相似性的纺织品图像增强

纺织品图像增强能够突出其纹理特性,便于纺织品的人工检测和机器视觉检测。提出一种在非局部均值滤波(NLM)框架下的纺织品图像纹理增强方法。纺织品图像具有规则周期的纹理,存在大量的冗余信息,NLM可利用这一特性来增强图像的纹理信息。但由于纺织品图像结构复杂且存在噪声,导致在NLM中相似性的度量不够准确。为解决这一问题,通过采用主分量分析(PCA)将纺织品图像分解为图像信息分量和噪声分量,并去除各分量间的相关性,来提高纺织品纹理间相似性度量的准确性。实验结果表明,本文方法比现有的纺织品图像纹理增强方法的增强效果有显著提高。     

Complexity, Confusion, and Perceptual Grouping. Part I: The Curve-like Representation

Intermediate-level vision is central to form perception, and we outline an approach to intermediate-level segmentation based on complexity analysis. We focus on the problem of edge detection, and how edge elements might be grouped together. This is typical because, once the local structure is established, the transition to global structure must be effected and context is critical. To illustrate, consider an edge element inferred from an unknown image. Is this local edge part of a long curve, or part of a texture? If the former, which is the next element along the curve? If the latter, is the texture like a hair pattern, in which nearby elements are oriented similarly, or like a spaghetti pattern, in which they are not? Are there other natural possibilities? Such questions raise issues of dimensionality, since curves are 1-D and textures are 2-D, and also of complexity. Working toward a measure of representational complexity for vision, in this first of a pair of papers we develop a foundation based on geometric measure theory. The main result concerns the distribution of tangents in space and in orientation, which serves as a formal basis for the concrete measure of representational complexity developed in the companion paper.     

Periodic pattern of texture analysis and synthesis based on texels distribution

Recently, sample-based texture synthesis techniques have drawn significant attention from researchers. These existing approaches mainly use the Markov Random Field (MRF) or texture features as texture model to analyze the local properties of sample textures. Indeed, human perception is sensitive to structure and periodicity. In this paper, we perform texture synthesis by taking into account the distribution of texels. Given a sample texture, the analysis procedure consists in segmenting texture into individual texels, and detecting each texel in order to analyze their neighborhood relationships by constructing connectivity. Then the synthesis process consists in reproducing a new large texture directly on a user-specified canvas by recomposing segmented texels, which synthesizes two-dimensional texel arrangements based on the previously constructed neighborhood relationships of texels. Results show that the proposed method is successful in generating textures visually indistinguishable to the sample textures. Moreover, the method especially deals with the near-regular textures, which well preserves underlying structural regularity.     

Computing oriented texture fields

The first step is the analysis of oriented texture consists of the extraction of an orientation field. The orientation field is comprised of the angle and coherence images, which describe at each point the dominant local orientation and degree of anisotropy, respectively. A new algorithm for computing the orientation field for a flow-like texture is presented. The basic idea behind the algorithm is to use an oriented filter, namely the gradient of Gaussian, and perform manipulations on the resulting gradient vector field. The most important aspect of the new algorithm is that it is provably optimal in estimating the local orientation of an oriented texture. An added strength of the algorithm is that it is simpler and has a better signal-to-noise ratio than previous approaches, because it employs fewer derivative operations. We also propose a new measure of coherence, which works better than previous measures. The estimates for orientation and coherence are related to measures in the statistical theory of directional data. We advocate the use of the angle and coherence images as intrinsic images. An analysis of oriented textures will require the computation of these intrinsic images as a first step. In this sense, the computation of the orientation field, resulting in the intrinsic images, is indispensible in the analysis of oriented textures. We provide results from several experiments to indicate the usefulness of the angle and coherence intrinsic images. These results show that the notion of scale plays an important role in the interpretation of textures. Further, measures defined on these intrinsic images are useful for the inspection of surfaces.     

基于主视通路结构分级响应模型的轮廓检测方法EI北大核心CSCD

基于视通路结构分级响应与动态传递的方式,本文提出了一种图像轮廓检测的新方法.针对视网膜感光细胞的暗视觉特性,建立亮度自适应的暗视野调节模型,利用多尺度经典感受野的方位选择性,构建高级轮廓与全局轮廓的检测路径;模拟外侧膝状体(Lateral geniculate nucleus,LGN)细胞特性对信息进行纹理稀疏编码,并结合非经典感受野的侧抑制作用抑制背景强纹理;另外在LGN区提出微动整合机制,减少纹理冗余信息,再经适应性突触实现信息关联传递;最后将初级轮廓响应跨视区前馈至V1区并经全局轮廓修正后,与高级轮廓响应实现快速融合.分别以RuG40、BSDS500图像库中的自然图像作为实验数据,检测结果与基准轮廓图的平均最优P指标分别为0.50、0.32,结果表明本方法能更有效地区分轮廓与纹理边缘,凸显主体轮廓.本文利用视神经细胞的内在机制以及神经信息的动态传递过程实现图像轮廓信息的编码与检测,也为研究后续高级视皮层的视觉感知提供了新思路.     

基于混合基稀疏图像表示的压缩传感图像重构

单一基函数不能对同时包含边缘和纹理信息的自然图像进行最优压缩传感图像重构. 本文根据Meyer的卡通--纹理图像模型和生物视觉原理, 用拉普拉斯塔式分解和圆对称轮廓波分别表示图像的光滑成分和边缘成分, 并构造了窄带轮廓波变换实现纹理成分的稀疏表示. 三种稀疏变换的基函数分别与视觉皮层中的侧膝体、简单细胞及栅格细胞的感受野类似. 结合三种图像稀疏表示方法和凸集交替投影算法提出了基于混合基稀疏表示的压缩传感图像重构算法. 实验结果表明,与基于块匹配三维变换迭代收缩的图像重构算法比较, 本文算法能获得更高的图像重构质量.     

Unsupervised detection and localization of structural textures using projection profiles

The main goal of existing approaches for structural texture analysis has been the identification of repeating texture primitives and their placement patterns in images containing a single type of texture. We describe a novel unsupervised method for simultaneous detection and localization of multiple structural texture areas along with estimates of their orientations and scales in real images. First, multi-scale isotropic filters are used to enhance the potential texton locations. Then, regularity of the textons is quantified in terms of the periodicity of projection profiles of filter responses within sliding windows at multiple orientations. Next, a regularity index is computed for each pixel as the maximum regularity score together with its orientation and scale. Finally, thresholding of this regularity index produces accurate localization of structural textures in images containing different kinds of textures as well as non-textured areas. Experiments using three different data sets show the effectiveness of the proposed method in complex scenes.     

A human vision based computational model for chromatic texturesegregation

We have developed a computational model for texture perception which has physiological relevance and correlates well with human performance. The model attempts to simulate the visual processing characteristics by incorporating mechanisms tuned to detect luminance-polarity, orientation, spatial frequency and color, which are characteristic features of any textural image. We obtained a very good correlation between the model's simulation results and data from psychophysical experiments with a systematically selected set of visual stimuli with texture patterns defined by spatial variations in color, luminance, and orientation. In addition, the model predicts correctly texture segregation performance with key benchmarks and natural textures. This represents a first effort to incorporate chromatic signals in texture segregation models of psychophysical relevance, most of which have treated grey-level images so far. Another novel feature of the model is the extension or the concept of spatial double opponency to domains beyond color, such as orientation and spatial frequency. The model has potential applications in the areas of image processing, machine vision and pattern recognition, and scientific visualization.     

Analysis and Controlled Synthesis of Inhomogeneous Textures

Many interesting real‐world textures are inhomogeneous and/or anisotropic. An inhomogeneous texture is one where various visual properties exhibit significant changes across the texture's spatial domain. Examples include perceptible changes in surface color, lighting, local texture pattern and/or its apparent scale, and weathering effects, which may vary abruptly, or in a continuous fashion. An anisotropic texture is one where the local patterns exhibit a preferred orientation, which also may vary across the spatial domain. While many example‐based texture synthesis methods can be highly effective when synthesizing uniform (stationary) isotropic textures, synthesizing highly non‐uniform textures, or ones with spatially varying orientation, is a considerably more challenging task, which so far has remained underexplored. In this paper, we propose a new method for automatic analysis and controlled synthesis of such textures. Given an input texture exemplar, our method generates a source guidance map comprising: (i) a scalar progression channel that attempts to capture the low frequency spatial changes in color, lighting, and local pattern combined, and (ii) a direction field that captures the local dominant orientation of the texture. Having augmented the texture exemplar with this guidance map, users can exercise better control over the synthesized result by providing easily specified target guidance maps, which are used to constrain the synthesis process.     

A Wavelet-Based Approach for Analyzing Industrial Stochastic Textures With Applications

Several continuous manufacturing processes use stochastic texture images for quality control and monitoring. Large amounts of pictorial data are acquired, providing important information about both the materials produced and the manufacturing processes involved. However, it is often difficult to measure objectively the similarity among industrial stochastic images or to discriminate between texture images of stochastic materials with distinct properties. Nowadays, the degree of discrimination required by industrial processes often goes beyond the limits of human visual perception. This paper proposes to model this specific class of textures as colored noise and presents a new approach for multiresolution stochastic texture representation and discrimination in industry (e.g., nonwoven textiles and paper). The wavelet transform is used to represent stochastic texture images in multiple resolutions and to describe them using local orientation and density variability as features. Based on this representation, a multiresolution distance measure for stochastic textures is proposed, and industrial applications of the method and experimental results are reported. The conclusions include ideas for future work     

Perception-based fuzzy sets for visual texture modelling

Texture is one of the most used low-level features for image analysis and, in addition, one of the most difficult to characterize. Although there is not an accurate definition for the concept of texture, it is usual for humans to describe visual textures according to some perceptual properties like coarseness, directionality, contrast, line-likeness or regularity. In this paper, we propose to model texture on the basis of its perceptual properties. To do this, fuzzy sets defined on the domain of some of the most representative measures of each property are employed. This approach achieves a double objective: first, to obtain models that allow to represent the imprecision related to texture properties, and second, to identify the most appropriate measure for each of these properties. In order to define the fuzzy models, parametric membership functions are proposed, where the corresponding parameters are obtained by learning a functional relationship between the computational values given by the measure and the human perception of the corresponding property. The performance of each fuzzy set is analyzed and checked with the human assessments, and a ranking of measures is obtained according to their ability to represent the perception of the property, allowing to identify the most suitable measure. In order to explain the proposed methodology, we focus our study on coarseness, contrast and directionality, that are considered the three most important texture properties.     

Factors Involved in Tactile Texture Perception Through Probes

An understanding of texture perception by robotic systems can be gained by examining human texture perception through a probe. As with texture perception in direct touch with the finger, texture perception by indirect touch by means of a probe is multidimensional, comprising rough, hard and sticky texture continua. In this study we describe variability among individual subjects in probe-mediated texture perception, and compare similarities and differences of texture perception between direct touch and indirect touch. The results show variability among subjects, as individual subjects may choose to rely on different texture dimensions, and examine the texture surface using different scanning velocities and contact forces. Despite the fact that the subjects use different scanning velocities and forces, they discriminate textures reliably and make subjective magnitude estimates consistently along a texture continuum when indirectly exploring texture surfaces with a probe. These data will be potentially useful in designing robotic sensory systems and understanding of sensory feedback, which is essential to teleoperations.     

A Mathematical Model of Retinal Ganglion Cells and Its Applications in Image Representation

Contrast stimuli are the essential components that constitute visual scenes. This paper studies how retinal ganglion cells respond to a contrast stimulus covering their concentric receptive fields with center-surround antagonism, and thereby proposes a mathematical model that describes the response by the multiplication of the contrast of the stimulus and a normalized response function with respect to the coverage ratio and the center/surround ratio. The obtained response curves turn out to be consistent with physiological data. This model partially accounts for the contrast sensitivities of ganglion cells and the invariance of visual perception to the contrast. A computational approach based on this neural model is developed for orientation detection, and is further applied to image representation. The experiments achieve convincing results on challenging image datasets. Moreover, it is revealed that the produced orientation maps remarkably enhance the efficiencies and the effectiveness of segmentation algorithms.     

一种方向Gabor滤波纹理分割算法

结合人眼视觉特性，设计了一种方向Gabor滤波器，该滤波器顾及了纹理图像的方向特性；利用Gabor滤波器的带通技术，抑制次要纹理图像的主频率分量，增强目标纹理图像主频率分量，使滤波输出图像具有较大的类间离散度和较小的类内离散度，将纹理图像的分割转化为传统的图像分割，使图像的分割质量和算法效率都得到了提高。     

Simulating Wood Using a Voxel Approach

In this paper we present a technique for generating three-dimensional wood textures using a regular texture array. Currently three-dimensional wood textures are generated using procedural textures. Procedural textures are flexible and require little memory, however the modeling of local artifacts such as knots is difficult using the procedural approach. By representing the wood as a texture array and growing the wood in this array we can easily simulate local phenomena such as knots. Our growth model is an approximation to the biological model and assumes that there are several similar wood cells per array element. This means that we can model artifacts that are defined by groups of similar cells. In particular our model is well suited for the modeling of soft-woods.     

Texturing fluids

We present a novel technique for synthesizing textures over dynamically changing fluid surfaces. We use both image textures as well as bump maps as example inputs. Image textures can enhance the rendering of the fluid by either imparting realistic appearance to it or by stylizing it, whereas bump maps enable the generation of complex micro-structures on the surface of the fluid that may be very difficult to synthesize using simulation. To generate temporally coherent textures over a fluid sequence, we transport texture information, i.e. color and local orientation, between free surfaces of the fluid from one time step to the next. This is accomplished by extending the texture information from the first fluid surface to the 3D fluid domain, advecting this information within the fluid domain along the fluid velocity field for one time step, and interpolating it back onto the second surface -- this operation, in part, uses a novel vector advection technique for transporting orientation vectors. We then refine the transported texture by performing texture synthesis over the second surface using our "surface texture optimization" algorithm, which keeps the synthesized texture visually similar to the input texture and temporally coherent with the transported one. We demonstrate our novel algorithm for texture synthesis on dynamically evolving fluid surfaces in several challenging scenarios.     

多水平外区抑制的轮廓检测

提取自然图像中的物体轮廓是机器视觉研究的重要问题,主要困难在于自然图像中的纹理性边缘严重干扰了物体轮廓的提取。研究表明视皮层方位选择性神经元的非经典感受野机制使得人类视觉系统在处理自然图像时不仅能够抑制纹理性边缘,而且能够增强物体的轮廓。基于此人们提出多种仿生轮廓检测算法,但算法中被称为抑制水平的参量在取值较高时会漏检部分轮廓,而在其取值较低时又会引入过多的纹理性边缘。针对这一问题,提出多水平外区抑制轮廓检测算法,通过整合各级单水平外区抑制的检测信息,有效抑制了纹理性边缘和降低了漏检轮廓的可能性。实验结果表明,相对于传统算法,新算法在轮廓检测性能上提高了10%左右,并具有更好的稳健性。     

The Promise and Perils of Near-Regular Texture

Motivated by the low structural fidelity for near-regular textures in current texture synthesis algorithms, we propose and implement an alternative texture synthesis method for near-regular texture. We view such textures as statistical departures from regular patterns and argue that a thorough understanding of their structures in terms of their translation symmetries can enhance existing methods of texture synthesis. We demonstrate the perils of texture synthesis for near-regular texture and the promise of faithfully preserving the regularity as well as the randomness in a near-regular texture sample.