TextureSynthesis

Research.TextureSynthesis History

Hide minor edits - Show changes to markup

January 24, 2008 by 134.226.85.142 -
Changed lines 10-16 from:
to:
April 26, 2007 by 134.226.86.54 -
Changed lines 10-13 from:
to:
February 26, 2007 by 134.226.86.238 -
Changed line 14 from:
to:
February 23, 2007 by 134.226.86.54 -
Changed lines 79-80 from:
  1. Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.
  2. Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.
to:
  1. Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.
  2. Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.
February 23, 2007 by 134.226.86.54 -
Changed line 78 from:
  1. Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.
to:
  1. Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.
February 23, 2007 by 134.226.86.54 -
Changed lines 78-84 from:
to:
  1. Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.
  2. Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.
  3. Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.
  4. Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in Irish Machine Vision and Image Processing (IMVIP), Dublin, Sep 2004.
February 23, 2007 by 134.226.86.54 -
Changed lines 78-87 from:

Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in Irish Machine Vision and Image Processing (IMVIP), Dublin, Sep 2004.

to:
February 23, 2007 by 134.226.86.238 -
Changed lines 7-8 from:

Below are some examples of synthesised texture generated using a novel wavelet based texture synthesis algorithm. The example texture is shown in the middle inside the black square and the new synthesised texture is grown around this example texture ``seed''. This algorithm combines the benefits of non-parametric modelling with wavelet based texture analysis. Unlike previous non-parametric approaches the algorithm is scale robust and generates impressive results for a wide range of textures. In addition, because most of the computationally intensive analysis work is performed at the coarse resolution, the algorithm is computationally efficient making it suitable for generating large texture images.

to:

Below are some examples of synthesised texture generated using a novel wavelet based texture synthesis algorithm. The example texture is shown in the middle inside the black square and the new synthesised texture is grown around this example texture "seed". This algorithm combines the benefits of non-parametric modelling with wavelet based texture analysis. Unlike previous non-parametric approaches the algorithm is scale robust and generates impressive results for a wide range of textures. In addition, because most of the computationally intensive analysis work is performed at the coarse resolution, the algorithm is computationally efficient making it suitable for generating large texture images.

February 23, 2007 by 134.226.86.238 -
Changed lines 5-6 from:

The problem of texture synthesis is a large research area in the computer graphics industry and in image post production and has received much attention in recent years. Given an example texture as a small sub-image, the idea behind a successful texture synthesis algorithm is to create a new (typically larger) image by generating or \textit{synthesising) more texture. This new synthesised texture should be perceptually similar and thus give the impression of being generated from the same underlying statistical process as the example texture.

to:

The problem of texture synthesis is a large research area in the computer graphics industry and in image post production and has received much attention in recent years. Given an example texture as a small sub-image, the idea behind a successful texture synthesis algorithm is to create a new (typically larger) image by generating or synthesising more texture. This new synthesised texture should be perceptually similar and thus give the impression of being generated from the same underlying statistical process as the example texture.

February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
Changed lines 78-83 from:

Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

to:

Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

February 23, 2007 by 134.226.86.238 -
Changed lines 80-81 from:

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

to:

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
Changed lines 80-81 from:

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

to:

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

February 23, 2007 by 134.226.86.238 -
Changed lines 18-19 from:
to:

To further demonstrate the strength of this new approach, an extensive comparison has been done between it and results obtained using a wide variety of previous approaches. The original texture images are taken from the Brodatz collection and the synthesised results obtained using the various approaches are shown below.

February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
Changed lines 25-26 from:
  • Parametric Approaches
to:
  • Our Results
Changed line 28 from:
to:
Changed lines 31-33 from:
  • Non-Parametric Approaches
  • Patch Based Approaches
to:
  • Kokaram (2000)
  • Portilla and Simoncelli (2000)
  • Efros and Leung (1999)
  • Wei and Levoy (2000)
  • Ashikhmin (2001)
  • Hertzmann (2001)
  • Efros and Freeman (2001)
February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
Added line 18:
Added line 20:
Added line 24:
Added lines 27-30:
February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
Changed lines 19-23 from:
  • Parametric Approaches''
to:
  • Parametric Approaches
February 23, 2007 by 134.226.86.238 -
Deleted lines 14-15:
Changed lines 17-22 from:
to:

Comparision with Previous Approaches

  • Original Images
  • Parametric Approaches''
  • Non-Parametric Approaches
  • Patch Based Approaches
February 23, 2007 by 134.226.86.238 -
Changed lines 10-13 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 14-16 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 10-11 from:
\\
\\
to:
Changed lines 14-16 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 10-14 from:
to:
\\
\\
February 23, 2007 by 134.226.86.238 -
Changed lines 17-18 from:

Comparison with Previous Approaches

to:
February 23, 2007 by 134.226.86.238 -
Added lines 17-18:

Comparison with Previous Approaches

February 23, 2007 by 134.226.86.238 -
Changed lines 20-25 from:

Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

to:

Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

February 23, 2007 by 134.226.86.238 -
Added lines 19-25:

Example Based Image Processing, C. Gallagher, PhD thesis, University of Dublin, Trinity College, Oct 2006.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

Changed lines 28-31 from:

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

to:
February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
February 23, 2007 by 134.226.86.238 -
Changed lines 12-14 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 12-14 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 10-12 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 10-12 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 10-18 from:
to:
February 23, 2007 by 134.226.86.238 -
Changed lines 12-15 from:
to:
February 22, 2007 by 134.226.86.54 -
Changed lines 10-23 from:
to:
February 22, 2007 by 134.226.86.238 -
Changed lines 10-16 from:
to:
February 22, 2007 by 134.226.86.238 -
February 22, 2007 by 134.226.86.238 -
Added lines 14-16:
February 22, 2007 by 134.226.86.238 -
Changed lines 10-17 from:
to:
February 22, 2007 by 134.226.86.54 -
Changed lines 10-12 from:
to:
February 22, 2007 by 134.226.86.238 -
Changed lines 3-4 from:

Keywords: Non-parametric Texture Synthesis, Post-Production, Wavelet

to:

Keywords: Texture Synthesis, Image modelling, Wavelet Based Analysis.

February 22, 2007 by 134.226.86.238 -
Changed lines 7-8 from:

Below are some examples of synthesised texture generated using a novel wavelet based texture synthesis algorithm. The example texture is shown in the middle inside the black square and the new synthesised texture is grown around this example texture ``seed''. This algorithm combines the benefits of non-parametric modelling with wavelet based texture analysis. Unlike previous non-parametric approaches the algorithm is scale robust and therefore works for a wide range of textures. In addition, becuase most of the computationally intensive analysis work is performed at the coarse resolution, the algorithm is fast.

to:

Below are some examples of synthesised texture generated using a novel wavelet based texture synthesis algorithm. The example texture is shown in the middle inside the black square and the new synthesised texture is grown around this example texture ``seed''. This algorithm combines the benefits of non-parametric modelling with wavelet based texture analysis. Unlike previous non-parametric approaches the algorithm is scale robust and generates impressive results for a wide range of textures. In addition, because most of the computationally intensive analysis work is performed at the coarse resolution, the algorithm is computationally efficient making it suitable for generating large texture images.

February 22, 2007 by 134.226.86.238 -
Changed lines 21-22 from:
to:

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher, F. Kelly and A. Kokaram, in IEE Conference on Vision Media Production (CVMP), London, Dec 2005.

February 22, 2007 by 134.226.86.238 -
Changed lines 17-19 from:
  • Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in Irish Machine Vision and Image Processing (IMVIP), Dublin, Sep 2004
to:

Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in Irish Machine Vision and Image Processing (IMVIP), Dublin, Sep 2004.

Non-Parametric Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in IEEE International Conference on Image Processing (ICIP), Geneva, Sep 2005.

February 22, 2007 by 134.226.86.238 -
Changed lines 5-6 from:

Texture synthesis is an important process in image post production. The idea behind a texture synthesis algorithm is that: given a small input sample of texture, create or synthesis a much later texture image which will be perceived by humans to be the same texture. Two different approaches to this problem have emerged. Parametric methods attempt to model the image with some definable process. These are efficient but do not work well on complicated textures. Non-parametric methods are simpler and rather attempt to measure the probability density function of each individual pixel using just the sample image itself. This approach was popularised by Efros and Leung in 1999. Because of the wide variability in image behaviour, non-parametric approaches have achieved by far the most pleasing results. However, one of the downsides to non-parametric methods is their computational costs and their dependence on scale. To address these problems we have developed a new algorithm which uses wavelet decomposition as a basis for non-parametric texture synthesis. Below are some of the results obtained using the algorithm. A full description of the algorithm is given here.

to:

The problem of texture synthesis is a large research area in the computer graphics industry and in image post production and has received much attention in recent years. Given an example texture as a small sub-image, the idea behind a successful texture synthesis algorithm is to create a new (typically larger) image by generating or \textit{synthesising) more texture. This new synthesised texture should be perceptually similar and thus give the impression of being generated from the same underlying statistical process as the example texture.

Below are some examples of synthesised texture generated using a novel wavelet based texture synthesis algorithm. The example texture is shown in the middle inside the black square and the new synthesised texture is grown around this example texture ``seed''. This algorithm combines the benefits of non-parametric modelling with wavelet based texture analysis. Unlike previous non-parametric approaches the algorithm is scale robust and therefore works for a wide range of textures. In addition, becuase most of the computationally intensive analysis work is performed at the coarse resolution, the algorithm is fast.

February 07, 2007 by 193.189.66.74 -
Changed lines 1-2 from:

by Claire GallagherKeywords: Non-parametric Texture Synthesis, Post-Production, Wavelet

to:

by Claire Gallagher

Keywords: Non-parametric Texture Synthesis, Post-Production, Wavelet

February 07, 2007 by 193.189.66.74 -
Changed lines 15-19 from:

1. N. Harte and A. Kokaram (2006) Automated Removal of Overshoot Artefact from Images. In 2006 European Signal Processing Conference (EUSIPCO 2006).. (BibTeX) 2. D. Corrigan , N. Harte and A. Kokaram (2006) Pathological Motion Detection for Robust Missing Data Treatment in Degraded Archived Media. In IEEE ICIP. Atlanta, USA, pages 621-624. (BibTeX) 3. D. Corrigan , A. Kokaram , R. Coudray and B. Besserer (2006) Robust Global Motion Estimation From Mpeg Streamswith a Gradient Based Refinement. In IEEE ICASSP. Toulouse, France, pages 285-288. (BibTeX) 4. Daire Lennon , Naomi Harte , Anil Kokaram , ErikaDoyle and Ray Fuller (2006) A HMM Framework for Motion based parsing for video fromObservational Psychology. In IEEE Irish Machine Vision and Image Processing Conference. September. ((URL)?) (BibTeX)

to:
February 07, 2007 by 193.189.66.74 -
Changed lines 6-8 from:
 alt=synthtised"
to:
Deleted lines 10-17:

|<img class="pic" src="simg5.jpg" alt="" width="150px" /> <img class="pic" src="timg5.jpg" alt="synthtised" width="300px"/> <br class="break"/> |<img class="pic" src="simg6.jpg" alt="" width="150px" /> <img class="pic" src="timg6.jpg" alt="synthtised" width="300px"/> <br class="break"/> |<img class="pic" src="simg1.jpg" alt="" width="150px" /> <img class="pic" src="timg1.jpg" alt="synthtised" width="300px"/> <br class="break"/> |<img class="pic" src="simg2.jpg" alt="" width="150px" /> <img class="pic" src="timg2.jpg" alt="synthtised" width="300px"/> <br class="break"/>

|<img class="pic" src="simg3.jpg" alt="" width="150px" /> <img class="pic" src="timg3.jpg" alt="synthtised" width="300px"/> <br class="break"/> |</center>

February 07, 2007 by 193.189.66.74 -
Changed lines 5-6 from:

|<center> |<img class="pic" src="simg4.jpg" alt="" width="150px" /> <img class="pic" src="timg4.jpg" alt="synthtised" width="300px"/> <br class="break"/>

to:
 alt=synthtised"
February 01, 2007 by 134.226.85.142 -
Changed lines 19-23 from:
   1. N. Harte and A. Kokaram (2006) Automated Removal of Overshoot Artefact from Images. In 2006 European Signal Processing Conference (EUSIPCO 2006).. (BibTeX)
   2. D. Corrigan , N. Harte and A. Kokaram (2006) Pathological Motion Detection for Robust Missing Data Treatment in Degraded Archived Media. In IEEE ICIP. Atlanta, USA, pages 621-624. (BibTeX)
   3. D. Corrigan , A. Kokaram , R. Coudray and B. Besserer (2006) Robust Global Motion Estimation From Mpeg Streamswith a Gradient Based Refinement. In IEEE ICASSP. Toulouse, France, pages 285-288. (BibTeX)
   4. Daire Lennon , Naomi Harte , Anil Kokaram , ErikaDoyle and Ray Fuller (2006) A HMM Framework for Motion based parsing for video fromObservational Psychology. In IEEE Irish Machine Vision and Image Processing Conference. September. ((URL)?) (BibTeX) 
to:

1. N. Harte and A. Kokaram (2006) Automated Removal of Overshoot Artefact from Images. In 2006 European Signal Processing Conference (EUSIPCO 2006).. (BibTeX) 2. D. Corrigan , N. Harte and A. Kokaram (2006) Pathological Motion Detection for Robust Missing Data Treatment in Degraded Archived Media. In IEEE ICIP. Atlanta, USA, pages 621-624. (BibTeX) 3. D. Corrigan , A. Kokaram , R. Coudray and B. Besserer (2006) Robust Global Motion Estimation From Mpeg Streamswith a Gradient Based Refinement. In IEEE ICASSP. Toulouse, France, pages 285-288. (BibTeX) 4. Daire Lennon , Naomi Harte , Anil Kokaram , ErikaDoyle and Ray Fuller (2006) A HMM Framework for Motion based parsing for video fromObservational Psychology. In IEEE Irish Machine Vision and Image Processing Conference. September. ((URL)?) (BibTeX)

February 01, 2007 by 134.226.85.142 -
Added lines 18-23:
   1. N. Harte and A. Kokaram (2006) Automated Removal of Overshoot Artefact from Images. In 2006 European Signal Processing Conference (EUSIPCO 2006).. (BibTeX)
   2. D. Corrigan , N. Harte and A. Kokaram (2006) Pathological Motion Detection for Robust Missing Data Treatment in Degraded Archived Media. In IEEE ICIP. Atlanta, USA, pages 621-624. (BibTeX)
   3. D. Corrigan , A. Kokaram , R. Coudray and B. Besserer (2006) Robust Global Motion Estimation From Mpeg Streamswith a Gradient Based Refinement. In IEEE ICASSP. Toulouse, France, pages 285-288. (BibTeX)
   4. Daire Lennon , Naomi Harte , Anil Kokaram , ErikaDoyle and Ray Fuller (2006) A HMM Framework for Motion based parsing for video fromObservational Psychology. In IEEE Irish Machine Vision and Image Processing Conference. September. ((URL)?) (BibTeX) 
January 21, 2007 by 134.226.86.54 -
Changed lines 3-4 from:

Texture synthesis is an important process in image post production. The idea behind a texture synthesis algorithm is that: given a small input sample of texture, create or synthesis a much later texture image which will be perceived by humans to be the same texture. Two different approaches to this problem have emerged. Parametric methods attempt to model the image with some definable process. These are efficient but do not work well on complicated textures. Non-parametric methods are simpler and rather attempt to measure the probability density function of each individual pixel using just the sample image itself. This approach was popularised by Efros and Leung in 1999. Because of the wide variability in image behaviour, non-parametric approaches have achieved by far the most pleasing results. However, one of the downsides to non-parametric methods is their computational costs and their dependence on scale. To address these problems we have developed a new algorithm which uses wavelet decomposition as a basis for non-parametric texture synthesis. Below are some of the results obtained using the algorithm. A full description of the algorithm is given here.

to:

Texture synthesis is an important process in image post production. The idea behind a texture synthesis algorithm is that: given a small input sample of texture, create or synthesis a much later texture image which will be perceived by humans to be the same texture. Two different approaches to this problem have emerged. Parametric methods attempt to model the image with some definable process. These are efficient but do not work well on complicated textures. Non-parametric methods are simpler and rather attempt to measure the probability density function of each individual pixel using just the sample image itself. This approach was popularised by Efros and Leung in 1999. Because of the wide variability in image behaviour, non-parametric approaches have achieved by far the most pleasing results. However, one of the downsides to non-parametric methods is their computational costs and their dependence on scale. To address these problems we have developed a new algorithm which uses wavelet decomposition as a basis for non-parametric texture synthesis. Below are some of the results obtained using the algorithm. A full description of the algorithm is given here.

November 16, 2006 by 127.0.0.1 -
Deleted line 0:
Changed lines 5-11 from:

http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg4.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg4.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg5.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg5.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg6.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg6.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg1.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg1.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg2.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg2.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg3.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg3.jpg

to:

|<center> |<img class="pic" src="simg4.jpg" alt="" width="150px" /> <img class="pic" src="timg4.jpg" alt="synthtised" width="300px"/> <br class="break"/> |<img class="pic" src="simg5.jpg" alt="" width="150px" /> <img class="pic" src="timg5.jpg" alt="synthtised" width="300px"/> <br class="break"/> |<img class="pic" src="simg6.jpg" alt="" width="150px" /> <img class="pic" src="timg6.jpg" alt="synthtised" width="300px"/> <br class="break"/> |<img class="pic" src="simg1.jpg" alt="" width="150px" /> <img class="pic" src="timg1.jpg" alt="synthtised" width="300px"/> <br class="break"/> |<img class="pic" src="simg2.jpg" alt="" width="150px" /> <img class="pic" src="timg2.jpg" alt="synthtised" width="300px"/> <br class="break"/>

|<img class="pic" src="simg3.jpg" alt="" width="150px" /> <img class="pic" src="timg3.jpg" alt="synthtised" width="300px"/> <br class="break"/> |</center>

November 16, 2006 by 127.0.0.1 -
Added lines 1-15:

by Claire GallagherKeywords: Non-parametric Texture Synthesis, Post-Production, Wavelet

Texture synthesis is an important process in image post production. The idea behind a texture synthesis algorithm is that: given a small input sample of texture, create or synthesis a much later texture image which will be perceived by humans to be the same texture. Two different approaches to this problem have emerged. Parametric methods attempt to model the image with some definable process. These are efficient but do not work well on complicated textures. Non-parametric methods are simpler and rather attempt to measure the probability density function of each individual pixel using just the sample image itself. This approach was popularised by Efros and Leung in 1999. Because of the wide variability in image behaviour, non-parametric approaches have achieved by far the most pleasing results. However, one of the downsides to non-parametric methods is their computational costs and their dependence on scale. To address these problems we have developed a new algorithm which uses wavelet decomposition as a basis for non-parametric texture synthesis. Below are some of the results obtained using the algorithm. A full description of the algorithm is given here.

http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg4.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg4.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg5.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg5.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg6.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg6.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg1.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg1.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg2.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg2.jpg
http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/simg3.jpg http://www.mee.tcd.ie/~sigmedia/research/DigitalCinema/texture/timg3.jpg

Bibliography

  • Wavelet Based Textured Synthesis, C. Gallagher and A. Kokaram, in Irish Machine Vision and Image Processing (IMVIP), Dublin, Sep 2004
Page last modified on January 24, 2008