bayesFMM2.htm0100755024222001473450000001251407561063232016350 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Bayes Fast Marching.

    tiger(256x179) P(O|X)= 4 P(O|X)= 29
bib.htm0100755024222001473450000003756207561070023015365 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

References

Application Examples

  • [Giak2000] "Digital Restoration of Painting Cracks.", Ioannis Giakoumis and Ioannis Pitas, IEEE on Image Processing February 2000.
  • [Chet2002] "Finding defects in texture using regularity and local orientation.", D. Chetverikov and
  • [Lew2001] "Lifting Detail from Darkness", J.P.Lewis , ACM Siggraph 2001 Technical Sketch.

General Papers

  • Decision Systems.
    • [Scho1994] "Logique floue et régulation PID.", Henk Scholten, Publitronic 1994.
    • [Pedr1993] "Fuzzy Control and Fuzzy Systems.", Witold Pedrycz, Research Studies Press, J. Wiley 1993.
    • [Boni1992] "Attribute Fusion and Situation Assessment with a many-Valued Logic Approach. Multi-Target MultiSensor Tracking. Application Advances.", R.M.Yannone and P.P.Bonissone,Yaakov Bar-Shalom, Artech 1992.
  • Classification.
    • [Ther1989] "Decision, Estimation and Classification.", Charles W. Therrien, Naval Postgraduate School, Wiley 1989.
    • [Datt1994] "Numerical Linear Algebra and Applications.", Biswa Nath Datta, Brools/cole Publishing Company 1994.
  • Evolution and Fast Marching.
    • [Seth1985] "Curvature and the Evolution of Fronts.", Sethian , Report 1985, New York University.
    • [Seth1995] "A fast marching level set method for monotonically advancing fronts.", Sethian , Proceedings 1995, UCLA.
    • [cohe1996] "Global minimum for active contour models: A minimal path approach.", Cohen & Kimmel , Proceedings IEEE 1996, San Francisco.
    • [Adal1997] "The fast construction of extension velocities in level sets methods.", Adalsteinsson & Sethian , report 1997, UCLA.
    • [Kimm2001] "Optimal Algorithm for Shape from Shading and Path Planning.", Ron Kimmel, James A. Sethian, Kluweer Academic Publishers 2001 , Netherlands.

Restoration Papers

    A large number of restoration techniques will be used in this framework. They will help people to recover images of different kind. Also it is wise that people will select themselve the suitable restoration method:

  • General papers.
    • [Blak1987] "Visual Reconstruction.", A. Blake and A. Zisserman , The MIT Press, Cambridge, Massachusetts Institute of Technology 1987, London, England.
    • [Marq1999] "Explicit algorithms for a new time dependent model based on level set motion for nonlinear deblurring an noise removal.", Marquina & S. Osher , Report 1999, UCLA.
    • [Mraz199x] "Nonlinear Diffusion of Image Derivatives.", P. Mrazek , Report 199x, Czech Technical University.
  • Fast Recursive Filters and Implicit Restorations.
    • [Deri1993] "Recursively Implementing the Gaussian and Its Derivatives.", R. Deriche , Report 1893, 1993, INRIA, France.
    • [DeBr1999] "Recursive Digital Filters.",Victor De Brunner , John Wiley Encyclopedia of Electrical and Electronic Engineer 1999, Volume 18.
    • [Kim199x] "Project: Heat transfert with ADI and PCG-ILU.",Dr. Seongjai Kim , Dep of Mathematics, University of Kentucky, Lexington.
    • [Lai993] "An O(N) Iterative Solution to the Poisson Equation in Low-level Vision Problems.", Lai and Vemuri, Technical Report TR-93-035, 1993, Universty of Florida.
    • [Weic1998] "Efficient and Reliable Schemes for Nonlinear Diffusion Filtering.", Joachim Weickert, report 1998, Utrecht University Hospital.
    • [Weic1999] "Fast Parallel Algorithms for a Broad Class of Nonlinear Variational Diffusion Approaches.", Joachim Weickert & all, report 1999, University of Mannheim, Germany.
    • [Weic2000] "Efficient Image Segmentation Using Partial Differential Equations and Morphology.", Joachim Weickert & all, technical report February 2000, University of Mannheim, Germany.
  • Specialized diffusions.
    • [Mala1995] "Level Set methods for curvature flow, Image enhancement and shape recovery in medical imaging.", R. Malladi & J.A. Sethian, proceeding berlin 1995 , UCLA.
    • [Blom1997] "Color TV: Total Variation Methods for Restoration of Vector Valued Images.", Peter Blomgren & Tony Chan, report 1997 , UCLA.
    • [Kimm1998] "Image processing via beltrami operator.", Ron Kimmel, R. Malladi & all, report 1998 , UCLA/Technion.
    • [Korn1998] "Contribution à la restauration d'images et à l'analyse de séquences: Approches Variationnelles et Solutions de Viscosité.", P. Kornprobst & all,Inria 1998.
    • [Chan2000] "Digital TV filter and nonlinear denoising.", Tony Chan , Stanley Osher & Jianhong Shen, report 2000 , UCLA.
    • [Weic1998] "Coherence-Enhancing Diffusion of colour images.", Joachim Weickert, report 1998, Utrecht University Hospital.
    • [Tang1998] "Direction Diffusion.", Bei Tang, report 1999, University of Minnesota.
    • [Weic2000] "A Scheme for Coherence-Enhancing Diffusion Filtering with Optimize Rotation Invariance.", Joachim Weickert & H. Scharr, Computer Vision 2000.
    • [Leve2000] "Level Set Based segmentation with intensity and curvature priors.", Michael E. Leventon, O. Faugeras & all, MIT & Inria 2000.
  • Enhancement.
    • [Coul2000] "Dual Echo MR Image Processing using multi-spectral probabilistic coupled with shock filters.", O. Coulon & all, report 2000, University College of London.
    • [Tchu2001] "Constrained and unconstrained PDE's for vector image restoration.", D. Tchumperlé & Rachid Deriche, Inria 2001.
  • Wavelets, Lifting and Restoration.
    • [Uytt1997] "The Red Black Transform.", G. Uytterhoeven A. Bultheel, Report TW271 1997, Katholieke Universiteit Leuven, Belgium.
    • [Holm1999] "Solving Hyperbolic PDEs using Intrerpolating Wavelets.", Mats HolmStrom, SIAM 1999.
    • [Vasi2001] "Solving Multi-Dimensionnal Evolution Problems with LocalizedStructures Using Second Generation Wavelets.", Oleg V. Vasilyev, Report University Missouri-Columbia 2001.

Inpainting Papers

    • [Masn1998] "Level lines based disocclusion.", Simon Masnou & JM Morel, Report 1998, Ceremade Université de Paris-IX Dauphine, France.
    • [Masn1998] "Image Restoration involving connectedness.", Simon Masnou & JM Morel, Report 1998, Ceremade Université de Paris-IX Dauphine, France.
    • [Cao1998] "A Generalization of Absolutely Minimizing Lipschitz Extension.", Frederic Cao, Report June 1998, Ecole Normale Supérieure de Cachan, France.
    • [Cass1998] "Image Interpolation.", Vicent Caselles & all, Report July 1998, University of Palma de Mallorca, Spain.
    • [Cass1998] "An Axiomatic Appoach to Image Interpolation.", Vicent Caselles & all, Report July 1998, University of Palma de Mallorca, Spain.
    • [Bert2000] "Image Inpainting.", Marcelo Bertalmio & all, Report 2000, University of Pompeu Fabra, Spain.
    • [Oliv2001] "Fast Digital Image Inpainting.", Manuel M.Oliveira & all, Proceedings on the International Conference on Visualization, Marbella 2001, Spain.
    • [Ball2000] "A Variational Model for Filling-In.", C. Ballester, Marcelo Bertalmio & all, Report 2000, University of Pompeu Fabra, Spain.
    • [Ball2001] "Filling-In by Joint Interpolation of Vector Fields and Gray Levels.", C. Ballester, Marcelo Bertalmio & all, Report April 2001, University of Pompeu Fabra, Spain.
    • [Bert2001] "Navier-Stokes, Fluid Dynamics, and Image and Vidéo Inpainting.", Marcelo Bertalmio, Report 2001, University of Pompeu Fabra, Spain.
    • [Bert2001] "Processing of Flat and non-flat image information on arbitrary manifolds using Partial Differential Equations.", Marcelo Bertalmio, Thesis March 2001, University of Minnesota.
    • [Chan1999] "Mathematical Models for Local Non-Texture Inpaintings.", Tony F. Chan & J. Shen, Report UCLA 1999.
    • [Chan2000] "Non-Texture Inpainting by Curvature-Driven Diffusion (CDD).", Tony F. Chan & J. Shen, Report UCLA 2000.
    • [Chan2001] "Euler'sElastica and Curvature Based Inpaintings.", Tony F. Chan & J. Shen, Report UCLA 2001.
    • [Chan2001] "Morphological Invariant PDE Inpaintings.", Tony F. Chan & J. Shen, Report UCLA 2001.
    • [Chan2001] "Digital Inpainting Based On the Mumford-Shah-Euler Image Model.", Tony F. Chan & J. Shen, Report UCLA 2001.
    • [Wei2000] "Fast Texture Synthesis using Tree-structured Vector Quantization.", L-Yi Wei & Marc Levoy, Report Stanford 2000.
    • [Perl199x] "B-spline Wavelet Paint.", Ken Perlin & Luis Velho, Report New York 199x.
calciteBeltrami.htm0100755024222001473450000001363207532306474017717 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

TV Filtering.(Scale 50%)

    Original image(512x350) Beltrami, time= 58 s Geodesic Beltrami, 100 iter, time= 58 s
    Kimmel ROF, 100 iter, time= 78 s Mean Curvature, 100 iter, time= 53 s TV Beltrami, 100 iter, time= 126 s
calciteBlurring.htm0100755024222001473450000001362707532306474017750 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Blurring and soft images.(Scale 50%)

    Original image(512x350) IRR sigma=3, time= 1.7 s IRR sigma=10, time= 1.7 s
    Mean Curvature, 100 iter, time= 53 s Sethian/ROF, 100 iter, time= 53 s Sethian/Beltrami, 100 iter, time= 65 s
calciteColorAnis.htm0100755024222001473450000001345407532306474020053 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

TV Filtering.(Scale 50%)

    Original image(512x350) Color Shock, iter= 50, time= 65 s
    ColorShapiro, 100 iter, time= 145 s Selective Deriche, 100 iter, time= 145 s Combined Deriche, 100 iter, time= 233 s
calciteEnhancement.htm0100755024222001473450000001351207532306474020402 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Contrast and Enhancement.(Scale 50%)

    Original image(512x350) Morphologic contrast, time= 4.2 s
    Simple Shock, 50 iterations, time= 36 s Selective Simple Shock, 50 iter, time= 56 s Color Shock, 50 iterations, time= 65 s
calcite.htm0100755024222001473450000001402007532310361016215 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Calcite(Zwitzerland)

    Calcite Crystal(512x350) Mean Curvature, 100 iter, time= 53 s
    Beltrami, 100 iter, time= 58 s Rudin/Osher/Fathemi-Sethian, 100 iter, time= 75s
    Color Shock(more contrast), 50 iter, time= 65 s Combined deriche (anisotropic diffusion+shock), 100 iter, time= 233 s
calciteTV.htm0100755024222001473450000001341107532306474016504 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

TV Filtering.(Scale 50%)

    Original image(512x350) TV, lambda=0.1, time= 82 s TV, lambda dynamic, time= 116 s
    TV gauss-jacobi, 100 iter, time= 117 s TV 4th Lysaker, 100 iter, time= 130 s
cinamon_titanite.htm0100755024222001473450000001274307532306474020162 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Cinamon

    Cinamon of Yunnan(512x350)
    Color Shock(more contrast) 30 iter Color Shock(more contrast) followed by beltrami, 50 iter
doc.htm0100755024222001473450000002237610050412477015373 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Information.

Java Active documentation in Portable Applets

The project features new active documentation of C++ classes in pure Java, running in a browser, with no additional installations.

This permits people wanting a fly overview, like deciders and consultants, to make rapid presentations, avoiding implementation deep analysis. Leaving potential client or coworkers with a first good impression of capabilities, before deciding deeper project investments.

In case o developpers, I write the Java code very near the C++ developped, to reduce study time, and increase understanding. I hope.The new Applets will also benefit of more information, parameter info, reading, ...

Java 1.1x is used to reduce all portable problems related to higher releases, giving most people their chances. We know more can be featured with higher releases, but this is not our aim here.

C++ Restoration & Inpainting Class Library (needs an update)

Crack detection

    One of the best crack detector is the top hat transform, which is not sensitive to light variation. We were able to restore old painting of the XVIs, by detecting cracks, inpainting and fast restorations.

Illumination and Color Enhancement

    Improving the image brightness without changing color uses simple technology like the histogram equalisation, gamma correction. The result is a greyish image loosing colors. One of the reason is that illumination in the image is non uniform.

    The Retinex Theory shows that the image can be separated in two parts the illuminance and the reflectance. Knowing the illuminance allows to do more appropriate gamma correction.

    In Homomorphic filters the illuminance is approximated by a luminance blurring. Kimmel's Variational Retinex allows to compute more accurate illuminance, taking into account visual properties and illuminance constraints. The result is a better color rendering during gamma modification. Finally, selecting an appropriate color space, can improve even more the results, avoiding chrominance losses.

Restoration

    Accurate restoration which maintains image details is important. We use sophisticated techniques based on the Alvarez-Lions-Morlet restoration model, to recover as much as possible the original image.

    Such techniques use, for example, complex differential equations which will reduce noise through diffusion, and enhance images through shocks.

    Some of these techniques may require a large amount of time(1 to 6 minutes, PIII/500Mhz). This is do to the difficulty of the restoration problem. We will progressively speedup these techniques when possible. We will also add alternative fast restorations.

    Fast AOS restoration reduces the time to 0.4 to 15 seconds(PIII/500Mhz) (divided be 2.5 with SSE1)

Inpainting

    Inpainting is an advanced interpolation scheme. It can be derived from differential equations and Sethian Level Sets...

    In short, the image is an hot area of pixels, and the cracks in the image,or unknown lost parts, are cold pixels. By maintaining the temperature of the pixels of the image, we will warm up the crack image areas on the basis of a border modulated conductivity factor which can locally change on a "material" basis.(e.g.: a metal becomes faster hot than wood,...). In our case the material is the image pixels, edges, curvature,...

    One of the simplest and fastest inpainting scheme, comes from Oliveira, it uses simple diffusion masks, well suited for removing thin areas.

    Faster techniques may exist, in practical condition, a weighted variant of the fast marching does a good and very fast job on thin objects.

    Inpainting on large areas is still a challenging difficult job. Mutiresolution facilitates inpainting. It permits to look images from far, and move to the image. Details are first small, and steadily will grow in size. This will happen to the inpainting areas too.

example1.htm0100755024222001473450000001632207560055471016345 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Restoration Examples

Inpainting Examples

Segmentation Examples (reduced size)

Tools

If your images are to dark, your monitor has a low gamma. We provide a gamma corrector in the software. On PC(24 bits displays are lighter than 32 bits)
exampleAltRestoration.htm0100755024222001473450000001415407622277652021166 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Mean Shift versus Fast Restoration.

The Mean Shift competes very well against Fast Restorations, in very noisy situations.
Fast Restorations take less time, and are more effective when we do not want to flatten the images.
    Original image(1024x518). : The Titanic taken from an original picture in a Belgium newspaper, December 1912.
    AOS Perona Fast Restoration(RGB): 4 iterations, tau=5, sigma=1, contrast=1, time=36s, with SSE1 9.28s
    Mean Shift with space sigma=8, range sigma=8, time=109s
    Mean Shift with space sigma=16, range sigma=16, time=443s
exampleBayes.htm0100755024222001473450000001333407545665557017267 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Color Bayes Classification Bi-Object/Multi-Class.

    Fountain Object Pass Filter
    Pont d'Avignon Object Pass Filter foreground
exampleBayesMine.htm0100755024222001473450000001436007561546101020055 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Color Bayes Classification Bi-Objects (background/foreground).

    Calcite Crystal ObjectFilter Foreground Label
    Blende (Spain) ObjectFilter Foreground Label
    Benitoite/Joachinite(San Benito, US) ObjectFilter Foreground Label
exampleBayesMulti.htm0100755024222001473450000001342607561545454020273 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Color Bayes Classification Bi-Object/Multi-Class.

    Benitoite(blue) and Joachinite(orange) (San Benito)
    ObjectFilter Multi ObjectFilter
    Foreground Label Multi Labels
exampleCalciteBench.htm0100755024222001473450000001313707532306474020513 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Restoration of Calcite Crystals.

Performance table

Example Restorations (images are of reduced size)

exampleCrackDetection1.htm0100755024222001473450000001365707570310236021152 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Test of Top Hat Transform and Cracks on Old Paintings Test Samples.

    Painting sample with cracks
    Crack detection using Black Top Hat (radial, closing size=8)
    Crack detection (thresholds: 32 and 92)
    Cracks
    Restored Painting using Fast Inpaintings
exampleCrackDetection2.htm0100755024222001473450000001325407570143574021155 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Test of Top Hat Transform and Cracks on Old Paintings Test Samples.

    Painting with two areas
    Crack detection
    Cracks
    Restored Painting using Fast Inpaintings
exampleCrackDetection.htm0100755024222001473450000001413607570143574021073 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Test of Top Hat Transform and Cracks on Old Paintings Test Samples.

    Painting sample with cracks
    Painting sample with cracks
    Restored Painting using Fast Inpaintings
    Painting sample with cracks
    Painting sample with cracks
    Restored Painting using Fast Inpaintings
exampleFastMarching.htm0100755024222001473450000001313107530511606020540 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Fast Marching and Incremental Distance Mapping.

    Original image Selective distance map, time= 60 ms
    Moving inside some pixels, time= less than 50 ms Moving inside with a known distance, time= less than 50 ms
exampleFluoriteBench.htm0100755024222001473450000001352107532306474020735 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Restoration of a Fluorite Crystal.

Performance table

Example Restorations (images are of reduced size)

exampleFMM.htm0100755024222001473450000001271607561062303016620 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Fast Marching Segmentation.

Example of Fast Marching Tracking

example.htm0100755024222001473450000002243610024072650016252 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Grey and Color Image Enhancement Examples

Restoration Examples

Inpainting Examples

Segmentation Examples (reduced size)

Tools

If your images are to dark, your monitor has a low gamma. We provide a gamma corrector in the software. On PC(24 bits displays are lighter than 32 bits)
exampleIIR.htm0100755024222001473450000001320507530511606016617 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Fast Recursive Filter(IIR) Blurring for Linear Time Scale Spaces.

    Original image Blurred image, sigma= 3, time= 440 ms
    Blurred image, sigma= 10, time= 440 ms Blurred image, sigma= 20, time= 440 ms, (small border artifacts of IIR)
exampleInpainting.htm0100755024222001473450000001212607530511606020275 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Inpainting.



  • Recovered with a combined inpainting, time= 60s.
exampleLifting.htm0100755024222001473450000001220307530511606017565 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

The Red Black Transform.


  • Original image.

  • Transformation with three levels.
exampleMedicalMR.htm0100755024222001473450000001304407560326534020001 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Restorations of MRI Brain Images.

Performance table

Example of Restorations

exampleMinerals.htm0100755024222001473450000001253207561544526017763 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Minerals.

Performance table

Example Restorations (images FULL size)

exampleMInpaintingSpace.htm0100755024222001473450000001303207530511606021363 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Multiresolution improves inpainting on large areas.

    Original image
    Simple Fast Oliveira cannot full fill the inpainting, time= 660 ms
    Multiresolution Fast Oliveira, time= 280 ms
exampleMInpaintingTime.htm0100755024222001473450000001335007530511606021231 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Multiresolution speeds up inpainting.

    Original image
    Simple Fast Oliveira, time= 2300 ms
    Multiresolution Fast Oliveira, time= 1270 ms
    One Shot Simple Fast Marching (works on homogeneous background), time= 170 ms !!!
example_old.htm0100755024222001473450000001400007530511606017103 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Picture Restoration.



Inpainting.



  • Recovered with a combined inpainting.


Fast Marching.

    Original image Selective distance map
    Moving inside some pixels Moving inside with a known distance
exampleRestoration.htm0100755024222001473450000001251607561062303020510 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Combined Picture Restoration.


  • Zeebrugge 1995, 400mm

    Zeebrugge restored with shock(50)+beltrami(100)

    Zeebrugge restored with shock(50)+ROF/sethian min-max(100)
exampleRetinex.htm0100755024222001473450000001524610024077220017610 0ustar bernarddecuyperrestoreinpaint Retinex and improving brightness with no color losses
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Retinex and improving brightness with no color losses

The Retinex is improving brightness of color images, by computation of the illumination image. A resulting gamma correction take care of light conditions. The color image maintain all his colors, this was not the case with classic gamma.

The implementation is based on Kimmel's Variational Retinex(Spie 2002). I augment the paper by adding an YIQ variant.

    Lena(384x384) in Color Gamma 1.6 Homomorphic filter+Gamma 1.6
    YIQ Retinex on Y channel HVS Retinex on V channel Histogram Equalisation
    RGB Retinex alpha=0.03, beta=0.25 gamma=1.6 RGB Retinex alpha=0.2, beta=0.0 saturates RGB Retinex alpha=0.0, beta=0.4 contrast
exampleRMSelective.htm0100755024222001473450000001330707560326534020370 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Multi Selective Restorations

    Bridge Left bottom tree
    Pont d'avignon
    Multiple selective restoration
exampleRSelective.htm0100755024222001473450000001373207560326534020255 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Single Selective Restorations

    Calcite Crystal(512x350) Mask
    Beltrami, 100 iter, time= 58 s Selective Beltrami, 100 iter, time= 9.67 s
    Rudin/Osher/Fathemi-Sethian, 100 iter, time= 75s Selective Rudin/Osher/Fathemi-Sethian, 100 iter, time= 19.5s
fluoriteBeltrami.htm0100755024222001473450000001360107532306474020140 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Beltrami.(FULL Scale)

    Fluorite Crystal(256x256) 25 iterations, time= 7 s 50 iterations, time= 14 s
    100 iterations, time= 24 s 200 iterations, time= 50 s 500 iterations, time= 128 s
fluoriteCDeriche.htm0100755024222001473450000001364007532312602020040 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Combined Deriche 2001.(FULL Scale)

    Fluorite Crystal(256x256) 25 iterations, time= 23 s 50 iterations, time= 46 s
    100 iterations, time= 90 s 200 iterations, time= 178 s 500 iterations, time= 450 s
fluoriteCShapiro.htm0100755024222001473450000001363407532306474020117 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Color Shapiro.(FULL Scale)

    Fluorite Crystal(256x256) 25 iterations, time= 11 s 50 iterations, time= 22 s
    100 iterations, time= 45 s 200 iterations, time= 89 s 500 iterations, time= 220 s
fluoriteEnhancement.htm0100755024222001473450000001321007532312200020602 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Contrast and Enhancement.(FULL Scale)

    Fluorite Crystal(256x256) Color Shock, 12 iter, time= 6 s
    Color Shock, 25 iter, time= 12 s Color Shock, 50 iter, time= 24 s
fluoritePerona.htm0100755024222001473450000001357207532306474017634 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Perona 1990.(FULL Scale)

    Fluorite Crystal(256x256) 25 iterations, time= 7 s 50 iterations, time= 14 s
    100 iterations, time= 27 s 200 iterations, time= 60 s 500 iterations, time= 150 s
fluoriteTV.htm0100755024222001473450000001336707532306474016743 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Simple TV(lambda=0.1).(FULL Scale)

    Fluorite Crystal(256x256) 50 iterations, time= 17 s
    100 iterations, time= 24 s 200 iterations, time= 70 s 500 iterations, time= 165 s
fluoriteWeickert.htm0100755024222001473450000001361107532306474020157 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Weickert 1998.(FULL Scale)

    Fluorite Crystal(256x256) 25 iterations, time= 15 s 50 iterations, time= 30 s
    100 iterations, time= 60 s 200 iterations, time= 119 s 500 iterations, time= 300 s
index1.html0100755024222001473450000002141707535725361016202 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

restoreInpaint

restoreInpaint is a state of the art project about making images "better". This can be done in different ways:

  • Restoration allows the recovery of an image, removing noise and enhancing details. Useful for paintings, holiday's and old paper pictures.
  • Inpainting covers the problem of detecting/filling cracks and missing parts of images, paintings, frescos.

Restoration

    A large number of restoration techniques will be used in this framework. They will help people to recover images of different kind. Also it is wise that people will select themselve the suitable restoration method:

    • People interested in general information, will blurr the image.
    • People looking for getting clean images, will go for getting smooth images.
    • People looking for getting clean images with clear borders, will smooth the images and maintain borders.
    • People requiring details will enhance details.
    • Combined operations are of course possible.

Inpainting

    The framework provides seven inpaintings with additional detection capabilities. People will be able to detect/remove thin artifacts, characters printed on pictures....

    • First, the user provides the inpainting area or uses an image crack detector.
    • With the known area, the inpainting will fill the gap using sophisticated techniques including:
      • Transportation used to move pixels from outside to inside cracks, taking account the details of the neigbour pixels.
      • Diffusion propagates your pixels from cracks borders, like something hot(the image) moves into a cold area(the crack area).
      • Multiresolution Looking to smaller images first will ease the inpainting. The results are propagated to larger scales images.

The Restoration & Inpainting C++ Class Library (Java style)

    This incremental library uses new technologies. Some of them are very fast, some are slow but with more restoration qualities(better edges,...). The technology is useful for learning purpose, explory purpose, research boosting, and application like old painting restoration, image recovery for medical or every day purpose,...

    • Spot and crack detection: Black and white Top Hat Transform.
    • Moving in image cracks: Sethian/Kimmel Fast Marching Method(also used in distance maps).
    • Multiresolution: Fast wavelet lifting and recursive filters(IIR)
    • Restoration: more than 30 restorations including fast lifting, fast IIR and recent partial differential equations.
    • Inpainting: more than 7 inpaintings using recent differential equations, including fast methods and multiresolution.
    • Morphology: grey level morphological operators: erosion, dilatation, opening, closing, contrast, top hat.
    • Math: Linear solvers Thomas, 7 Conjugate gradients. Planned: ADIs, preconditionner family.
    • Multiprocessing: A working thread manager, andd OO Image Workers can work together (forground/background , ...).
    • Segmentation: We are planning color bayes segmentation, Mean shift, fast marching tracking and derived level sets.
    • Acquisition: We are planning scanner acquisitions schemes.

    We show also how to improve and combine techniques together, to boost your imagination.

index.html0100755024222001473450000003317110050410554016076 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

restoreInpaint

restoreInpaint is a state of the art project about making 8 or 16bit depth images "better":

  • Detection covers the problem of finding target areas
  • Inpainting discovers the problem of filling detected cracks and missing thin parts of images, paintings, frescos.
  • Restoration allows the recovery of an image, removing noise, enhancing brightness, color and details, extraction and demixing. Usage: paintings, holiday's and old pictures..., Photobit pattern removal.
  • Complex Restorations with small theta can restore images and provide object boundaries at the same time.
  • Realtime and Fast Restorations are used with large images, like in medical imaging and astronomy. Details are restored via distance modulated restorations

What am I speeking about? Better, is to see an image, and develop freely your idea later.

    Original image (50% size) Crack removed & restored image

Restoration

    A large number of restoration techniques will be used in this framework. They will help people to recover images of different kind. Also it is wise that people will select themselve the suitable restoration method:

    • People interested in general information, will blur the image.
    • People looking for getting clean images, will go for getting smooth images.
    • People looking for getting clean images with clear borders, will smooth the images and maintain borders.
    • People requiring details will enhance details.
    • Deciders an consultants requiring direct on spot examples.Java in a browser presentations

Inpainting

    The framework provides seven inpaintings with additional detection capabilities. People will be able to detect/remove thin artifacts, characters printed on pictures....

    • First, the user provides the inpainting area or uses an image crack detector.
    • With the known area, the inpainting will fill the gap using sophisticated techniques including:
      • Transportation used to move pixels from outside to inside cracks, taking account the details of the neigbour pixels.
      • Diffusion propagates your pixels from cracks borders, like something hot(the image) moves into a cold area(the crack area).
      • Multiresolution Looking to smaller images first will ease the inpainting. The results are propagated to larger scales images.

The Restoration & Inpainting C++ Class Library (Java style)

    This incremental library uses new technologies. Some of them are very fast, some are slow but with more restoration qualities(better edges,...). The technology is useful for learning purpose, explory purpose, research boosting, and application like old painting restoration, image recovery for medical or every day purpose,...
    MinGW with OpenGL/Glut/GLUI and parallel MMX/SSE1. Maximum power, the cheapest solution .
    We also show a realtime restoration demo using MinGW with OpenPTC. Getting power of DirectX fast display.

    • Active documentation: Java Applet(1.15) are portable in most browsers, see explanation and real life results with a simple click , before moving to C++. The project features equivalent coding in JAVA and C++ for your best choice.
    • Image formats: 8bit grey and 24bit color for jpeg, bmp, TIFF. In version 1.1: PGM,PPM,PNM,PAM, including 16bit grey and 48 bit color TIFF and PGM/PPM (read/write)
    • Color Space Enhancement: Gamma, Histogram equalisation. New in 1.1d: Kimmel Variational Grey/RGB/YIQ/HVC Retinex, Homomorphic filter, image to image color transfert. RGB, XYZ, YIQ, Yuv, LHS, HSV, HVC, HSI, LMS, Lab, Luv. Coming: RLab
    • Restoration: more than 30 restorations using partial differential equations(explicit methods), alternative restorations Mean Shift, Barash bilateral filters, scharr coherence enhancement, Gilboa forward/backward diffusion, and complex diffusion.Planned complex shocks.
    • Fast Restoration: fast lifting, fast IIR and fast large time steps semi-implicit methods including 3 multiplicative splitting: LOD, Strang, AFI, and 7 additive splitting: AOS Perona variants, Pyramidal AOS, AOS FAB like Perona, NEW Barash AOS Enhancer/denoiser (in v1.1). In planning: multiresolution/superresolution versions.
    • Distance Fast Restoration: fast modulated time step fast restoration combined to the fast Marching including LOD for Heat Flow and AOS Perona .
    • Complex Fast Restoration: fast large time steps semi-implicit methods including LOD for Heat Flow and AOS Perona (v0.99,16/10/2003) .
    • Blind Source Separation: In preparation(v1.1)ICA and FastICA will give perspectives in feature detection, denoising and demixing images.
    • New Feature Extraction: In final testing(v1.1): KLT-PCA and Windowed PCA In preparation(v1.1)SVM will give perspectives in feature detection, image analysis.
    • Very Fast Restoration: using fast SSE Optimised Pentium III Alvarez-Deriche LOD blurrer, and Weickert's AOS Perona, FAB based AOS proto.
    • Multiresolution: Fast wavelet lifting and recursive filters(IIRS), gradient and laplacian IIRs, Gabor IIR
    • Superresolution: In 1.1, bilinear and bicubic interpolations, new interpolation including ENO interpolators coming.
    • Inpainting: more than 7 inpaintings using recent differential equations, including fast methods and multiresolution.
    • Spot and crack detection: Black and white Top Hat Transform. Including fast large SE processing.
    • Morphology: grey level and large morphological operators: erosion, dilatation, opening, closing, contrast, local threshold, top hat. MMX coding for grey images
    • Segmentation: color bayes segmentation, fast marching tracking.
    • Fast Marching: incremental distance maps, moving from border inside, border measurement analysis, path tracking(Non-linear and central), Bayes FMM.
    • Fast Sweeping: very fast iterative distance map generator, unlike FMM it builds distance maps in O(N), a serious advantage on large images. (more than 300% speed-up)
    • Math: Linear solvers Thomas, 7 Conjugate gradients. In finalisation(v1.1): QR family, pseudo-inverse, SVD Golub-Kahan, Lanczos, Inverse power, Oja, SPCA, APEX. New coming: Spared and block GS, SOR, Large number CG solvers including Truncated and Shift CG. Logistic for large scale linear solvers.
    • Differential and Accuracy: ENO2, ENO3, WENO3 non oscillating, and limiters like: minmod, superbee, mineno, harmod, hareno, variable superbee.
    • Multiprocessing: A working thread manager, and OO Image Workers can work together (foreground/background , ...).sqrt, 1/sqrt,1/x SSE speed up
    • Acquisition: scanner/digital camera acquisitions using Twain API.(Visual C++, not explored yet under MinGW)Planning: post-processing.

    We show also how to improve and combine techniques together, to boost your imagination.

license.htm0100755024222001473450000001263407560326534016256 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Copyrights: Eggs and Pictures 2002: Bernard De Cuyper, Eddy Fraiha.

Remember to put the above copyright line, in the first place of all sources, and displayed in the about box.

    Open BSD License Model.

    Eggs and Pictures is the organisation name under which we are working in team.

    Bernard De Cuyper and Eddy Fraiha are the effective team members of Eggs and Pictures.

    We left the license quite open to any usage type. The software is available for friendly usage.

links.htm0100755024222001473450000001602010103417325015727 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Links.

Organisations

Universities

Logistic

medicalR1.htm0100755024222001473450000001353107560326537016435 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

General restorations.

Most restorations use 100 iterations.
    Sagital View(256x256) Mean Curvature, time= 7 s TV Rudin/OSher/Fatemi, time= 9 s
    Beltrami, time= 7 s Classic Perona, time= 7 s Selective Shock(35iter), time= 5 s
medicalR2.htm0100755024222001473450000001263607560326537016443 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

TV restorations.

All restorations use 100 iterations.
    Sagital View(256x256) TV Rudin/OSher/Fatemi, time= 9 s TV of 4th order Lysaker, time= 9 s
medicalR3.htm0100755024222001473450000001426307560326537016442 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Restorations with first shock enhancement.

All restorations start with 20 iterations selective shocks, followed by 100 iterations main restorations.
    Sagital View(256x256) Classic Perona Alone Shock followed by Classic Perona Shock 20 iterations
    Beltrami Alone Shock followed by Beltrami TV Lysaker Order 4, Alone Shock followed by TV Lysaker Order 4
medicalR4.htm0100755024222001473450000001532707560330410016426 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

General restorations.

Main restorations use 100 iterations.
    Sagital View(256x256) Shock(20 iter)+ Regularised Perona Selective Shock(35iter)
    Vertico-frontal View(256x256) Shock(20 iter)+ Regularised Perona Selective Shock(35iter)
    Axial View(256x256) Shock(20 iter)+ Regularised Perona Selective Shock(35iter)
    Axial View(256x256) Shock(20 iter)+ Regularised Perona Selective Shock(35iter)
medicalR5.htm0100755024222001473450000001355307622277302016437 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Fast restorations.

.
    Sagital View(256x256) Mean Curvature(100 iter), time= 7 s Heat Flow step=5.0, time= 220 ms
    Classic Perona(100 iter), time= 7 s Alvarez/Perona step=2.5, time= 550 ms AOS Perona step=2.0, t=710 ms, with SSE1=110ms
news.htm0100644024310601473450000001006007506723624013522 0ustar benevrestoreinpaint New Page 1
News Downloads Documents Examples Bibliography Liscence Contact
ir.jpg (10573 bytes)
 

 

 

 

CONTENT HERE - LINK CELLS HIGHLIGHT.

 

 

 

 

pathtracking.htm0100755024222001473450000001253307734534652017316 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Fast Marching and Non linear Paths.

    Homogene euclidian path Non-linear image path Central path
pinkQuartz.htm0100755024222001473450000001414207561544151016776 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Pink quartz with maganapatite(Brazil)

    Pink quartz with maganapatite(512x350) Color Shock (increasing contrast), 50 iter
    Color Shock(more contrast) followed by beltrami, 50 iter Combined deriche (anisotropic diffusion+shock), 100 iter
    Sethian-beltrami(light smoothing and speckle denoise), 50 iter Sethian-Rudin/Osher/Fatemi(strong smoothing and speckle denoise), 50 iter
restoreComplex.htm0100644024222001473450000001605307744225171017642 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Complex restoration of images+boundaries.

(Time is computed on PIII/500Mhz)
  • Using Complex numbers with small theta in diffusion, allows the restoration at the same time of the image and the boundaries.(Gilboa2001)
  • The real part displays the restored image with less staircase effects that with real number restoration, but more diffuse results.
  • The complex part displays a laplacian restored image, scaled with time.
  • NEW LOD/AOS Fast complex restorations results are given.(Bernard De Cuyper, 15/10/2003, v0.99)
  • Thanks to Gilboa, Deriche and Weickert & all.
    Original Perona1990 (iter=75, 2.58s)
    Complex Linear Diffusion(Real)(iter=10, 2.86s) Complex Linear Diffusion(Img)(iter=10, 2.86s) Complex NonLinear Diff(Real)(iter=10, 4.25s) Complex NonLinear Diff(Img)(iter=10, 4.25s)
    Complex LOD heat flow(Real)(iter=1, 0.38s) Complex LOD heat flow(Img)(iter=1, 0.34s) Complex AOS NonLinear Diff(Real)(n=1, 1.11s) Complex AOS NonLinear Diff(Img)(n=1, 1.09s)
restoreFAB.htm0100644024222001473450000001440107741756727016633 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Forward/backward FAB restoration.

(Time is computed on PIII/500Mhz)
  • Using local blurring/deblurring in an anisotropic way, at the same time, improves restoration accuracy.(Gilboa2002)
  • The process shows that using shocks can be avoided, and may be less accurate.
  • We demonstrate an explicit FAB Perona, compared with the Perona1990 and Deriche selective shock.
  • A faster AOS Perona FAB implementation is also provided, with less accuracy, but serious improvements.
    Original Perona1990 (iter=75, 2.58s) Gilboa FAB Perona (iter=75, 4.75s)
    AOS SSE1 Perona (iter=5, 0.47s) AOS FAB Perona (iter=5, 0.92s) Selective Shock (iter=30, 2.30s)
selectiveRestoration.htm0100755024222001473450000001404107560326537021046 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Pink quartz with maganapatite(Brazil)

    Calcite Crystal(512x350) Mean Curvature, 100 iter, time= 53 s
    Beltrami, 100 iter, time= 58 s Rudin/Osher/Fathemi-Sethian, 100 iter, time= 75s
    Color Shock(more contrast), 50 iter, time= 65 s Combined deriche (anisotropic diffusion+shock), 100 iter, time= 233 s
similarFMM.htm0100755024222001473450000001344107561063037016626 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Fast Marching on central pixel value

    benitoite(256x179) difference= 12% difference= 40%
    difference= 46% difference= 54% difference= 57%
tourmaline.htm0100755024222001473450000001305307561062303016777 0ustar bernarddecuyperrestoreinpaint Image Restoration
Home Downloads SF Page Documentation Examples Bibliography Links License
SourceForge.net Logo

Blue Tourmaline on Quartz(Madagascar)

    Blue Tourmaline(512x350)
    Shock(20 iter) + Beltrami(100 iter)
    Shock(20 iter) + ROFSethian min/max(100 iter)