| authors: | A. D. Spence, M. J. Chantler |
| title: | On Capturing 3D Isotropic Surface Texture using Uncalibrated Photometric Stereo |
| abstract: | We propose an uncalibrated method of acquiring the normal and albedo fields of an isotropic 3D surface texture. The method is 'uncalibrated' in that it also provides estimates of the unknown illumination vectors. Thus, given a set of images taken from a single viewpoint but under varied and unknown illumination, our approach returns estimates of: the surface normal field, the albedo field, and the illumination vectors. We assume the use of single point lighting and Lambertian isotropic surfaces. We also assume that the major variations in the data are in the x-y plane - due to realistically rough surfaces. We use Hayakawa's uncalibrated photometric stereo algorithm to simultaneously estimate the scaled surface normals and the illumination vectors in an arbitrary co-ordinate system. As the variance in the data is mostly concentrated in the x-y plane, the z-axis is assumed to be given by the third eigen vector. Orientation in the x-y plane is determined by applying a frequency domain method for estimating illumination tilt angles. |
| keywords: | 3d surface texture ; ; Blank field ; |
| authors: | Cula, O.G. , Dana, K.J., Murphy F.P. & Rao. B.K. |
| title: | Bidirectional Imaging and Modeling of Skin Texture |
| abstract: | Human skin is a complex surface, with fine scale geometry and local optical properties that make its appearance difficult to model. Also, the conditions under which the skin surface is viewed and illuminated greatly affect its appearance. In this work, we capture the dependency of skin appearance on imaging parameters using bidirectional imaging. We construct a new skin texture database, containing bidirectional measurements of normal skin and of skin affected by various disorders. The complete database contains more than 3500 images, and is made publicly available for further research. Furthermore, we present two computational models for use in skin texture recognition. Both models are image-based representations of skin appearance that account for the varied appearance of the skin with changes in illumination and viewing direction. We employ these models in two contexts: discrimination between different skin disorders (e.g., psoriasis vs. acne), and classification of facial locations based on facial skin texture (e.g. forehead vs. chin). The classification experiments demonstrate the usefulness of the modeling and measurement methods. |
| keywords: | surface texture ; illumination effects & invariants ; viewpoint effects & invariants ; |
| authors: | Koudelka, M.L.; Magda, S.; Belhumeur, P.N.; Kriegman, D.J. |
| title: | Acquisition, Compression, and Synthesis of Bidirectional Texture Functions |
| abstract: | Real world surfaces such as tree bark, moss, sponge, and fur often have complicated geometry that leads to effects such as self-shadowing, masking, specularity, and interreflection as the lighting or viewpoint in a scene changes. We use image based techniques to analyze and represent bidirectional texture functions, or BTFs, with correct geometric and lighting effects. A basis for the apparent BRDF of points on the surface is determined and used to compress the texture datasets, as well as provide a space for comparison of texture elements across all lights and views. The compression method reduces the approximately 10,000 images in each 6-D lighting, viewpoint, and spatial variation texture dataset to just under 2MB. |
| keywords: | surface texture ; illumination effects & invariants ; BTF, viewpoint effects, compression ; |
| authors: | Spence, A.D. and Chantler, M.J. |
| title: | Optimal illumination for three-image photometric stereo acquisition of surface texture |
| abstract: | The optimal placement of the illumination for three-image photometric stereo when used for capturing 3D surface texture is derived and verified experimentally. The sensitivities of the scaled surface normal elements with respect to the input images are derived and used to provide expressions for the noise variances. An overall figure of merit is developed by considering image-based rendering (i.e. relighting) of Lambertian surfaces. This metric is optimised with respect to the illumination angles. The optimal difference between tilt angles of successive illumination vectors was found to be 120°. The optimal slant angle was found to be 90° for smooth surfaces and 55° for rough surfaces. |
| keywords: | surface texture ; ; Blank field ; |
| authors: | Turtinen M. & Pietikäinen M. |
| title: | Visual Training and Classification of Textured Outdoor Scene Images |
| abstract: | Classification of textures in scene images is very difficult due to the high variability of the data within and between images caused by effects such as non-homogeneity of the textures, changes in illumination, shadows, foreshortening and self-occlusion. For these reasons, finding proper features and representative training samples for a classifier is very problematic. Even defining the classes which can be discriminated with texture information is not so straightforward. In this paper, a visualization-based approach for training a texture classifier is presented. Powerful local binary patterns (LBP) are used as texture features and a self-organizing map (SOM) is employed for visual training and classification, providing very promising results in the classification of outdoor scene images. |
| keywords: | classification and/or segmentation ; surface texture ; visualization ; |
| authors: | Koki Fujita and Shree K. Nayar |
| title: | Recognition of Dynamic Textures using Impulse Responses of State Variables |
| abstract: | Dynamic textures are image sequences which contain moving scenes such as a flowing river, drifting smoke, waving foliage, etc. Such image sequences have dynamical properties that are related to motions in the physical world. In this paper, we propose a novel analytical tool for analyzing dynamic textures. The key idea is to exploit the properties of the impulse responses of the state variables computed using the previous algorithm for dynamic texture representation. It turns out that the fundamental dynamical properties of a dynamic texture are captured very efficiently by these impulse responses. We have used our approach to develop an algorithm for the recognition of local dynamic textures. This algorithm is significantly more efficient than previously proposed techniques that use distances between computed dynamic texture models in a non-linear space. We test the recognition accuracy of our algorithm using a variety of real-world dynamic textures. |
| keywords: | classification and/or segmentation ; texture models ; dynamic textures ; |
| authors: | Yang, F. & Lishman, R. |
| title: | Land Cover Change Detection Using Gabor Filter Texture |
| abstract: | We wish to detect land cover change for environment management. Gabor filters are used to correlate with original land cover images to derive texture information. This paper investigates a texture based image description in which the standardised MPEG-7 Homogeneous Texture Descriptors (HTD) of Gabor filters are used as the textural feature vector. Then this vector is input to a discriminant classifier using linear regression analysis. This paper presents the result of possible change detection of arable land. Experiment results show that the MPEG-7 texture descriptor gives an efficient and effective classification rate on land cover images. |
| keywords: | classification and/or segmentation ; ; Blank field ; |
| authors: | Yiming Wu, Kap Luk Chan, Yong Huang |
| title: | Image Texture Classification Based on Finite Gaussian Mixture Models |
| abstract: | A novel image texture classification method based on finite Gaussian mixture models of sub-band coefficients is proposed in this paper. In the method, an image texture is first decomposed into several sub-bands, then the marginal density distribution of each sub-band coefficients is approximated by Gaussian mixtures.The Gaussian component parameters are estimated by an "EM+MML" algorithm which performs parameter estimation and model selection automatically. The Earth Mover's Distance (EMD) is used to measure the distribution similarity based on the Gaussian components.Thus, classification can be done by matching the marginal density distributions. Extensive experiments show that the proposed method achieved an overall improved classification accuracy compared to nonparametric representation of sub-band distributions. |
| keywords: | classification and/or segmentation ; Blank field ; Blank field ; |
| authors: | Abbadeni, N. |
| title: | A New Similarity Matching Measure: Application to Texture-based Image Retrieval |
| abstract: | This paper addresses the fundamental issue of similarity in image databases. A new similarity model is introduced based on Gower's coefficient of similarity. This similarity model is flexible and can be declined in several versions: non-weighted, weighted and hierarchical versions. The similarity model is applied on textures by considering two content representation models: the well-known autoregressive model and a perceptual model based on perceptual features such as coarseness and directionality. Experimentations conducted with human subjects, showing interesting results, are presented. |
| keywords: | database retrieval ; texture models ; Blank field ; |
| authors: | Sonia Starik, Michael Werman |
| title: | Simulation of Rain in Videos |
| abstract: | This paper treats the addition of simulated rain to a video sequence. In order to derive a fast and simple algorithm for rain simulation that produces realistic results, we investigated visual properties of rainfall in videos, in terms of time and space. Assuming partial knowledge of the intrinsic camera parameters and user-defined parameters regarding rain intensity and velocity, we derive visual properties of the rain "strokes" in the video space and show how to use these strokes to modify the video to give a realistic impression of rain. |
| keywords: | rendering & visualisation ; texture synthesis ; rain simulation ; |
| authors: | Geusebroek, J.M. |
| title: | A Scale-Space Analysis of Multiplicative Texture Processes |
| abstract: | Gaussian Scale-space describes the local structure of images. This paper shows a stochastic analysis of the diffusion equation as put forward by Koenderink (1984) for regular images. Important classes of the stochastic process which are structurally described by the diffusion analysis include Brownian fractals, Markovian textures, and fragmentation processes. The analysis shows the diffusion coefficient to relate to the local autocorrelation function over the diffusion process. |
| keywords: | texture feature design & performance ; texture models ; Blank field ; |
| authors: | Efstathios Hadjidemetriou, Michael Grossberg, Shree Nayar |
| title: | Multiresolution Histograms and their Use for Texture Classification |
| abstract: | The histogram of image intensities is used extensively for the retrieval of images from visual databases. An obvious way to extend this feature is to compute the histograms of multiple resolutions of an image. This extension shares many desirable properties with the plain histogram including that they are both fast to compute, space efficient, invariant to rigid motions, and robust to noise. In addition, the histograms over multiple image resolutions directly encode texture information. We describe a simple yet novel matching algorithm based on this extension. We evaluate it against algorithms based on five widely used texture features. The comparison is performed on two different texture databases. We show that with our simple feature, we achieve or exceed the performance obtained with more complicated texture features. Further, we show our algorithm to be the most efficient, robust, and appropriate particularly for large texture databases. |
| keywords: | texture feature design & performance ; classification and/or segmentation ; database retrieval ; |
| authors: | Salvatella, A. & Vanrell, M & Villanueva, J.J. |
| title: | Texture Description based on Subtexture Components |
| abstract: | The growing of multimedia content has motivated the need of tools to do image browsing and annotation; several texture descriptors have been proposed, but the high degree of complexity textures can achieve has limited their success. In this paper the concept of subtexture is introduced in order to make the automatic description of a texture adaptable to its complexity degree. A subtexture component is defined by sets of blobs or emergent patterns that have similar simple features, and can be fully described by a 7-dimensional vector, similar to the PBC descriptor proposed in MPEG-7 standard. Thus, we propose a comprehensive texture description formed by the descriptions of its N subtexture components, that is, a Nx7 matrix where the number of rows is related to the complexity of the texture. In this work a multiscale method to identify the subtexture components is presented. It is based on automatic scale selection for blob detection. Once the subtextures are identified, a global feature analysis provides the attributes of each subtexture component. Finally, the comprehensive descriptor is built from combining all subtexture information. |
| keywords: | texture models ; ; subtexture, texture descriptor ; |
| authors: | Dongxiao Zhou and Georgy Gimel'farb |
| title: | Model-based Estimation of Texels and Placement Grids for Fast Realistic Texture Synthesis |
| abstract: | This paper describes the basic steps of a new technique, called bunch sampling, that enables the realistic synthesis of spatially homogeneous textures. A geometric shape of the bunch (acting as a texel) and spatial placement grid governing relative positions of the bunches are estimated from the training texture by using a generic Gibbs random field texture model with multiple pairwise pixel interactions. During the synthesis, the bunches are randomly sampled from the training texture and placed into the large-size goal image with due account of their spatial interdependence. |
| keywords: | texture synthesis ; ; Markov/Gibbs random field, block sampling, stochastic texture, regular mosaic ; |
| authors: | Doretto, G. & Soatto, S. |
| title: | Towards Plenoptic Dynamic Textures |
| abstract: | We present a technique to infer a model of the spatio-temporal statistics of a collection of images of dynamic scenes seen from a moving camera. We use a time-variant linear dynamical system to jointly model the statistics of the video signal and the moving vantage point. We propose three approaches to inference, the first based on the plenoptic function, the second based on interpolating linear dynamical models, the third based on approximating the scene as piecewise planar. For the last two approaches, we also illustrate the potential of the proposed techniques with a number of experiments. The resulting algorithms could be useful for video editing where the motion of the vantage point can be controlled interactively, as well as to perform stabilized synthetic generation of video sequences. |
| keywords: | texture synthesis ; texture models ; viewpoint effects, rendering & visualization, colour texture, 3D surface texture ; |
| authors: | Yacov Hel-Or, Tom Malzbender, and Dan Gelb |
| title: | Synthesis of Reflectance Function Textures from Examples |
| abstract: | We extend the machinery of existing texture synthesis methods to handle texture images where each pixel contains not only RGB values, but reflectance functions. Like conventional texture synthesis methods, we can use photographs of surface textures as examples to base synthesis from. However multiple photographs of the same surface are used to characterize the surface across lighting variation, and synthesis is based on these source images. Our approach performs synthesis directly in the space of reflectance functions and does not require any intermediate 3D reconstruction of the target surface. The resulting synthetic reflectance textures can be rendered in real-time with continuous control of lighting direction. |
| keywords: | texture synthesis ; surface texture ; Blank field ; |
| authors: | Yanxi LIU and Steve LIN |
| title: | Deformable Texture: the irregular-regular-irregular cycle |
| abstract: | Blank field Departures from a regular texture pattern can happen in many different dimensions. Previous work has been focusing on faithful texture synthesis for near-regular texture for textures departing along the color and intensity axes while the underlying geometric regularity is well preserved. Examples include brick walls, tiled floor and woven straw sheet. In this paper, we address the issues of texture synthesis for near-regular texture that is geometrically distorted. We propose a framework that treats irregular texture as a deformation from regular texture by first deducing a deformation field between the input irregular texture and its corresponding regular version. The novel view in this work is to treat the deformation field itself as a texture that is both visual and functional. As a result, we can handle faithful texture synthesis for a much larger variety of near-regular to irregular textures. |
| keywords: | texture synthesis ; texture models ; regularity, near-regularity, symmetry, deformation fields, spring-network ; |
| authors: | Neubeck, A., Zalesny, A., and Van Gool, L. |
| title: | Cut-primed smart copying |
| abstract: | Texture synthesis through so-called 'smart copying' requires a seamless meshing of texture subparts. Current systems first select subparts that seem to globally fit well and then optimize the cut between them on a rather local scale. This order of first selecting patches first and then caring about the seamless meshing reduces one's leeway in the choice of seamless cuts to a small zone of overlap between the patches. Therefore, even in the latest and smartest of smart copying approaches seams still tend to show up in the results. Here we present an approach that first looks for promising cuts, and uses these as the point of departure. It is shown that even a simple criterion for the quality of seams already supports high-quality smart copying and texture tessellation. |
| keywords: | texture synthesis ; Blank field ; textured tessellation ; |
| authors: | Haindl, M. Filip. J. |
| title: | Fast BTF Texture Modelling |
| abstract: | This paper presents a fast model-based algorithm for realistic multispectral BTF tecapable of direct implementation inside the graphical card processing unit. The algorithm starts with range map estimation of the BTF texture followed by the spectral and spatial factorization of an input multispectral texture image. Single orthogonal monospectral components are decomposed into a multi-resolution grid and each resolution data are independently modelled by their dedicated Gaussian Markov random field model (GMRF). We estimate an optimal contextual neighbourhood and parameters for each GMRF. Finally single synthesized monospectral texture pyramids are collapsed into the fine resolution images and using the inverse Karhunen-Loeve transformation we obtain the smooth multispectral texture. Both multispectral and range information is combined in a bump mapping filter of the rendering hardware. The presented model offers huge BTF texture compression ration which cannot be achieved by any other sampling-based BTF texture synthesis method. |
| keywords: | texture synthesis ; colour texture ; BTF textures, compression ; |
| authors: | Dong, J. & Chantler, M. |
| title: | Comparison of Five 3D Surface Texture Synthesis Methods |
| abstract: | We present and compare five approaches for synthesising and relighting real 3D surface textures. We adapted Efros’s texture quilting method and combined it with five different relighting representations, comprising: a set of three photometric images; surface gradient and albedo maps; polynomial texture maps; and two eigen based representations using 3 and 6 base images. We used twelve real textures to perform quantitative tests on the relighting methods in isolation. We developed a qualitative test for the assessment of the complete synthesis systems. Ten observers were asked to rank the images obtained from the five methods using five real textures. Statistical tests were applied to the rankings. Our conclusion is that the cheaper gradient and three-base-image eigen methods should be used in preference to the other methods; especially where the surfaces are Lambertian or near Lambertian. |
| keywords: | texture synthesis ; Blank field ; Blank field ; |
| authors: | Dmitry Chetverikov and Zsolt Janko |
| title: | Periodic Textures in Wide Baseline Stereo |
| abstract: | This study addresses the problem of finding correspondences for wide baselinestereo. Texture has traditionally been utilised as a single-image cue for 3D shape reconstruction (shape-from-texture); at the same time, its role in multiview scene reconstruction has been very limited. In stereo image matching, repetitive patterns are usually considered as disturbing factor since they tend to produce multiple peaks of correlation, which results in matching ambiguity. We argue that presence and proper analysis of distinct, compact periodic texture areas can facilitate wide baseline matching by providing periodic distinguished regions (PDRs) that efficiently constrain the search for correspondences. We demonstrate how PDRs can be used to find a few initial correspondences in a wide baseline stereo pair and to establish precise correspondences for building the epipolar geometry. Experimental results for various wide baseline stereo pairs are shown. |
| keywords: | viewpoint effects & invariants ; ; Wide baseline stereo ; |