A Compact Model for Viewpoint Dependent Texture Synthesis

Zalesny, Alexey; Gool, Luc J. Van

doi:10.1007/3-540-45296-6_9

Cited by 38 publications

(38 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The pairwise MRFs employed here are very similar to models that have been popular for a long time [10]; the higher-order MRFs follow the recently proposed Fields-of-Experts (FoE) approach [9]. Richer models of natural images have also been proposed on the basis of MRFs with multiple pairwise pixel interactions [11,12]. We are not following this approach here, but a comparison of the benefits of these approaches deserves further study.…”

Section: Learning Markov Random Field Models Of Imagesmentioning

confidence: 99%

Efficient Belief Propagation with Learned Higher-Order Markov Random Fields

Lan

Roth

Huttenlocher

et al. 2006

Lecture Notes in Computer Science

166

167

View full text Add to dashboard Cite

Abstract. Belief propagation (BP) has become widely used for low-level vision problems and various inference techniques have been proposed for loopy graphs. These methods typically rely on ad hoc spatial priors such as the Potts model. In this paper we investigate the use of learned models of image structure, and demonstrate the improvements obtained over previous ad hoc models for the image denoising problem. In particular, we show how both pairwise and higher-order Markov random fields with learned clique potentials capture rich image structures that better represent the properties of natural images. These models are learned using the recently proposed Fields-of-Experts framework. For such models, however, traditional BP is computationally expensive. Consequently we propose some approximation methods that make BP with learned potentials practical. In the case of pairwise models we propose a novel approximation of robust potentials using a finite family of quadratics. In the case of higher order MRFs, with 2 × 2 cliques, we use an adaptive state space to handle the increased complexity. Extensive experiments demonstrate the power of learned models, the benefits of higher-order MRFs and the practicality of BP for these problems with the use of simple principled approximations.

show abstract

Section: Learning Markov Random Field Models Of Imagesmentioning

confidence: 99%

Efficient Belief Propagation with Learned Higher-Order Markov Random Fields

Lan

Roth

Huttenlocher

et al. 2006

Lecture Notes in Computer Science

166

167

View full text Add to dashboard Cite

show abstract

“…Texture can be synthesized by reproducing joint statistics of the exemplar [e.g. Zalesny and Van Gool 2001].…”

Section: Related Workmentioning

confidence: 99%

Parallel controllable texture synthesis

Lefèbvre

Hoppe

2005

ACM SIGGRAPH 2005 Papers

114

151

View full text Add to dashboard Cite

Exemplar Synthesized deterministic windowsMultiscale randomness Spatial modulation Feature drag-and-drop Figure 1: Given a small exemplar image, our parallel synthesis algorithm computes windows of spatially deterministic texture from an infinite landscape in real-time. Synthesis variation is obtained using a novel jittering technique that enables several intuitive controls. AbstractWe present a texture synthesis scheme based on neighborhood matching, with contributions in two areas: parallelism and control. Our scheme defines an infinite, deterministic, aperiodic texture, from which windows can be computed in real-time on a GPU. We attain high-quality synthesis using a new analysis structure called the Gaussian stack, together with a coordinate upsampling step and a subpass correction approach. Texture variation is achieved by multiresolution jittering of exemplar coordinates. Combined with the local support of parallel synthesis, the jitter enables intuitive user controls including multiscale randomness, spatial modulation over both exemplar and output, feature dragand-drop, and periodicity constraints. We also introduce synthesis magnification, a fast method for amplifying coarse synthesis results to higher resolution.

show abstract

“…Zalesny and Van Gool [Zalesny and Gool 2001] demonstrate a method of synthesizing oriented textures for surfaces, but their approach does not model texture quality, and thus replaces the low resolution texture, rather than using it to guide the synthesis.…”

Section: Texture Synthesismentioning

confidence: 99%

Detail synthesis for image-based texturing

Ismert

Bala

Greenberg

2003

Proceedings of the 2003 Symposium on Interactive 3D Graphics - SI3D '03

View full text Add to dashboard Cite

Interactive and immersive environments which model the real world often use images of the environment to capture realistic visual complexity. Image-based modeling techniques permit the creation of visually interesting geometric models from photographs. These models are textured by resampling these images of the scene; we call this process image-based texturing. The problem with traditional image-based texturing is the poor quality of the extracted textures, which are often blurred or stretched.This paper introduces a novel technique to improve the quality of image-based texturing processes by introducing a physically-based metric that can be used to extend current texture synthesis methods. We propose a sampling-based metric of texture quality based on the Jacobian matrix of the imaging transform, which captures the interaction of the imaging system with the imaged environment. This metric suggests a physical interpretation of the multi-resolution image representations widely used in texture synthesis. Use of this metric enables synthesis of high spatial frequency detail into regions of an image-based model's textures where the imaging process captures only low frequency texture data. Given a small set of input images and a geometric model of the scene, this technique allows the creation of uniform, high-resolution textures. Our technique relieves the user of the burden of collection large numbers of images and increases the quality of user-driven image-based modeling systems. This improved quality is important in order to create compelling visual experiences in interactive environments. * e-mail: mccloud@graphics.cornell.edu † e-mail:kb@cs.cornell.edu ‡ e-mail:dpg@graphics. AbstractInteractive and immersive environments which model the real world often use images of the environment to capture realistic visual complexity. Image-based modeling techniques permit the creation of visually interesting geometric models from photographs. These models are textured by resampling these images of the scene; we call this process image-based texturing. The problem with traditional image-based texturing is the poor quality of the extracted textures, which are often blurred or stretched.This paper introduces a novel technique to improve the quality of image-based texturing processes by introducing a physically-based metric that can be used to extend current texture synthesis methods. We propose a sampling-based metric of texture quality based on the Jacobian matrix of the imaging transform, which captures the interaction of the imaging system with the imaged environment. This metric suggests a physical interpretation of the multi-resolution image representations widely used in texture synthesis. Use of this metric enables synthesis of high spatial frequency detail into regions of an image-based model's textures where the imaging process captures only low frequency texture data. Given a small set of input images and a geometric model of the scene, this technique allows the creation of uniform, high-resolution textures. Our tec...

show abstract

A Compact Model for Viewpoint Dependent Texture Synthesis

Cited by 38 publications

References 15 publications

Efficient Belief Propagation with Learned Higher-Order Markov Random Fields

Efficient Belief Propagation with Learned Higher-Order Markov Random Fields

Parallel controllable texture synthesis

Detail synthesis for image-based texturing

Contact Info

Product

Resources

About