Image Texture Synthesis

50 %
50 %
Information about Image Texture Synthesis
Science-Technology

Published on October 23, 2010

Author: ahmedajaz

Source: authorstream.com

IMAGE TEXTURE SYNTHESIS : IMAGE TEXTURE SYNTHESIS Submitted by- Ajaz Ahmed Siddiqui Mohd. Shahid Ameen Roll No. 7013802 Roll No. 7013815 All rights reserved © ahmedajaz 2010 Contents : Contents What is Image Texture Synthesis Structured and Stochastic Structures The Goal of Texture Synthesis Commonly used techniques Non-parametric Sampling Image Quilting Texture Transfer Conclusion References All rights reserved © ahmedajaz 2010 What is Image Texture Synthesis? : What is Image Texture Synthesis? According to Wikipedia, “Texture synthesis is the process of algorithmically constructing a large digital image from a small digital sample image by taking advantage of its structural content.” It is object of research to computer graphics and is used in many fields, amongst others digital image editing, 3D computer graphics and post-production of films. All rights reserved © ahmedajaz 2010 Structured and Stochastic textures : Structured and Stochastic textures *Structured textures: These textures look like somewhat regular patterns. An example of a structured texture is a stonewall or a floor tiled with paving stones. *Stochastic textures: Texture images of stochastic textures look like noise: colour dots that are randomly scattered over the image, barely specified by the attributes minimum and maximum brightness and average colour. Many textures look like stochastic textures when viewed from a distance. An example of a stochastic texture is roughcast. All rights reserved © ahmedajaz 2010 Slide 5: Structured textures Stochastic textures All rights reserved © ahmedajaz 2010 Slide 6: The Goal of Texture Synthesis Given a finite sample of some texture, the goal is to synthesize other samples from that same texture The sample needs to be "large enough“ True (infinite) texture generated image input image SYNTHESIS All rights reserved © ahmedajaz 2010 Slide 7: Commonly used techniques for Texture Synthesis:- Pixel-based texture synthesis The simplest and most successful general texture synthesis algorithms. They typically synthesize a texture in scan-line order by finding and copying pixels with the most similar local neighborhood as the synthetic texture. They are typically accelerated with some form of Approximate Nearest Neighbor method since the exhaustive search for the best pixel is somewhat slow. “Texture Synthesis by Non-parametric Sampling." Efros and Leung, ICCV, 1999 All rights reserved © ahmedajaz 2010 Slide 8: Patch-based texture synthesis Patch-based texture synthesis creates a new texture by copying and stitching together textures at various offsets. These algorithms tend to be more effective and faster than pixel-based texture synthesis methods. "Image Quilting." Efros and Freeman. SIGGRAPH 2001 All rights reserved © ahmedajaz 2010 Non-parametric sampling : Non-parametric sampling It is a process by which a new image made from an initial seed (given sample), one pixel at a time. Here, a statistical non-parametric model based on the assumption of spatial locality is used. This method is best suited for constrained synthesis problems (hole filling). All rights reserved © ahmedajaz 2010 Slide 10: Markov Random Field (MRF) is used for modelling texture. We assume that the probability distribution of brightness values for a pixel, given the brightness values of its spatial neighborhood, is independent of the rest of the image. The spatial neighborhood is modelled as a square window around the pixel. If the texture is presumed to be mainly regular at high spatial frequencies and mainly stochastic at low spatial frequencies, the size of the window should be on the scale of the biggest regular feature. The method : All rights reserved © ahmedajaz 2010 Slide 11: Here I is an image that is synthesized from texture sample image Ismp Ireal (finite texture). Let p I be a pixel and let N(p) I be a square image of width N centered at p. To synthesize the value of p we first construct an approximation to the conditional probability distribution P(p|N(p)) and then sample from it and let d(N1, N2) denote some perceptual distance between two patches. Based on our MRF model we assume that p is independent of I \ N(p) given N(p). If we define a set. Ω(p)={ N’ Ireal : d(N’, N(p))=0 } containing all occurrences of N(p) in Ireal then the conditional pdf of p can be estimated with histogram of all center pixel values in Ω(p). All rights reserved © ahmedajaz 2010 Slide 12: Assuming Markov property, compute P(p|N(p)) Building explicit probability tables infeasible Instead, let’s search the input image for all similar neighborhoods — that’s our histogram for p To synthesize p, just pick one match at random Input image Synthesizing a pixel non-parametric sampling N All rights reserved © ahmedajaz 2010 Limitations : Limitations In Non-parametric algorithm is its tendency for some textures to occasionally “slip” into a wrong part of the search space and start growing garbage or get locked onto one place and produce verbatim copies(exactly same) of the original Algorithm “slips” into a wrong part of the search space and starts growing garbage The algorithm is quite slow but efforts are there trying to make it more efficient. All rights reserved © ahmedajaz 2010 Its applications : Its applications A tool for solving several practical problems in computer vision, graphics, and image processing. This method is particularly versatile because it does not place any constraints on the shape of the synthesis region or the sampling region, making it ideal for constrained texture synthesis such as holefilling. All rights reserved © ahmedajaz 2010 Slide 15: Texture synthesis algorithm is applied The synthesis process fills in the black regions. Examples All rights reserved © ahmedajaz 2010 Slide 16: Results This algorithm produces good results for a wide range of textures. The only parameter set by the user is the width ‘N’ of the context window. This parameter appears to intuitively correspond to the human perception of randomness for most textures. The user has to look after only the value of ‘N’ which is width of the context window INPUT INPUT OUTPUT OUTPUT All rights reserved © ahmedajaz 2010 Image Quilting : Image Quilting A simple image-based method of generating novel visual appearance in which a new image is synthesized by stitching together small patches of existing images. The input image is divided into set SB of square blocks, say Bi . The tiling of square blocks in random order from SB, is done over the new image. However the new image we get is not as satisfying as it should be. So smoothing is done at the edges. All rights reserved © ahmedajaz 2010 Slide 18: input idea After smoothing result 1st Step: tile input image; pick random blocks and place them in random locations; Smooth edges All rights reserved © ahmedajaz 2010 Slide 19: In the first step, we still see that the result is not convincing, thus instead of picking a random block, we will search SB for such a block that by some measure agrees with its neighbors along the region of overlap. An error in the overlap region between a chosen block and other blocks is computed and the minimum is chosen to make the cut. All rights reserved © ahmedajaz 2010 Slide 20: The minimal cost path through the error surface is computed in the following manner. If B1 and B2 are two blocks that overlap along their vertical edge with the regions of overlap B1ovand B2ov, respectively, then the error surface is defined as Ei,j = ei,j + min(Ei-1,j-1,Ei-1,j,Ei-1,j+1) In the end, the minimum value of the last row in E will indicate the end of the minimal vertical path though the surface and one can trace back and find the path of the best cut. When there is both a vertical and a horizontal overlap, the minimal paths meet in the middle and the overall minimum is chosen for the cut. All rights reserved © ahmedajaz 2010 Slide 21: Input texture B1 B2 Random placement of blocks block All rights reserved © ahmedajaz 2010 Slide 22: overlapping blocks vertical boundary All rights reserved © ahmedajaz 2010 Slide 23: Our final output block Thus the process could be repeated on whole of the image and the desired image could be obtained. This synthesis process works on a wide range of input textures, but its particularly effective for semi structured textures which was difficult earlier. Results All rights reserved © ahmedajaz 2010 Some Examples:- : Some Examples:- All rights reserved © ahmedajaz 2010 Slide 25: All rights reserved © ahmedajaz 2010 Slide 26: All rights reserved © ahmedajaz 2010 Texture Transfer : Texture Transfer Image Quilting can further be used for texture transfer. Texture transfer is taking the texture from one object and “painting” it onto another object. Source texture Target image Slide 28: Texture transfer involves correspondence map of images, which is spatial map of some corresponding quantity over both the texture source image and a controlling target image. That quantity could include image intensity, blurred image intensity, local image orientation angles, or other derived quantities. All rights reserved © ahmedajaz 2010 Slide 29: Source texture Target image All rights reserved © ahmedajaz 2010 Slide 30: + = All rights reserved © ahmedajaz 2010 Slide 31: + = + = parmesan rice All rights reserved © ahmedajaz 2010 Conclusion : Conclusion Non-parametric Sampling:- Very simple Surprisingly good results Synthesis is easier than analysis! …but very slow Image Quilting No one-pixel-at-a-time! fast and very simple Results are good for most images All rights reserved © ahmedajaz 2010 References: : References: Texture Synthesis by Non-parametric Sampling Alexei A. Efros and Thomas K. Leung(Sept. 1999) Computer Science Division University of California, Berkeley , CA 94720-1776, U.S.A. Image Quilting for Texture Synthesis and Transfer Alexei A. Efros(University of California, Berkeley) William T. Freeman(Mitsubishi Electric Research Laboratories) http://en.wikipedia.org/ http://graphics.cs.cmu.edu.university-archive.org/ All rights reserved © ahmedajaz 2010

Add a comment

Related presentations

Related pages

Texture synthesis - Wikipedia, the free encyclopedia

Texture synthesis is the process of algorithmically constructing a large digital image from a small digital sample image by taking advantage of its ...
Read more

Texture Analysis and Synthesis - Computer Graphics at ...

Texture synthesis can also produce tileable images by properly handling the ... images] "Fast Texture Synthesis using Tree-structured Vector ...
Read more

Image Quilting for Texture Synthesis and Transfer

Image Quilting for Texture Synthesis and Transfer Alexei A. Efros1,2 William T. Freeman2 1University of California, Berkeley 2Mitsubishi Electric Research ...
Read more

Texture Synthesis Links

Texture Links Texture Synthesis. Popat, '93: Novel cluster-based probability model for texture synthesis, classification, and compression. Heeger and ...
Read more

Texture Synthesis - Carnegie Mellon Computer Graphics

Texture Synthesis In the past decade computer graphics experienced a wave of activity in the area of image-based rendering as researchers explored the idea ...
Read more

Image Quilting for Texture Synthesis and Transfer

Image Quilting for Texture Synthesis and Transfer Alexei A. Efros and Willian T. Freeman. We present a simple image-based method of generating novel visual ...
Read more

Image Texture Tools - logarithmic.net

Image Texture Tools. ... synthesis of texture from a sample, ... a noise removal method that is guided by a model of an image's texture, ...
Read more

Fast Texture Synthesis using Tree-structuredVector ...

Fast Texture Synthesis using Tree-structuredVector Quantization Li-Yi Wei Marc Levoy Stanford University Figure 1: Our texture generation process takes an ...
Read more

Texture Synthesis Using Image Quilting - Computer Science ...

Imagequilt.m Main routine for image quilting Mincut.m Computes the shortest path from one end of an image to another Ssd.m Computes sum of squared ...
Read more

DEPTH IMAGE BASED RENDERING WITH ADVANCED TEXTURE SYNTHESIS

978-1-4244-7493-6/10/$26.00 ©2010 ieee icme 2010 depth image based rendering with advanced texture synthesis p. ndjiki-nya, m. köppel, d. doshkov, h ...
Read more