Research Material

Segmentation [SFSBM] [MD2S] [SCKM]
Fusion of Label Fields [PRIF] [FCR]
Restoration-Deconvolution [SFRBR] [DECWAV] [NLREST]
Denoising [SIMPAR]
Reconstruction [DCT-EM]
Hyperspectral Compression [BCOCDM[M4ICD]
HDR Compression [PAGETM]
Vizualisation [OBEHS]



This page has been created as a service to the image processing community to encourage reproducible research

  • The term reproducible research (first proposed by J. Claerbout at Stanford Univ.) and refers to the idea that the ultimate product of research is the paper along with the full computational environment used to produce the results in the paper such as the code, data, etc. necessary for reproduction of the results and building upon the research.
  • Note that the code given in this web page is optimized for simplicity, not speed ! (and this has to be mentioned if you report timing results for comparisons). This code should NOT be used for any commercial purposes without direct consent of their author(s) (see copyrights). Please cite the paper if you use this code for your research work

  • SFSBM Algorithm

    thumbnail
    A non-stationary MRF model for image segmentation from a soft boundary map
    Max Mignotte
    Pattern Analysis & Aplications (to appear)
      Abstract: We address the problem of estimating a segmentation map into regions only from a soft (or possibly probabilistic) boundary representation. For this purpose, we have defined a nonstationary multiscale MRF model with long-range pairwise interactions whose potentials are estimated from the likelihood of the presence of an edge at each considered pair of pixels. Another contribution of this paper is also to demonstrate that an efficient alternative strategy to region-based segmentation models consists in averaging several (quickly estimated) soft contour maps and to exploit this MRF reconstruction model in order to achieve a reliable segmentation into regions.
      ___________
    • Segmentation results on the Berkeley image database obtained by our algorithm (with param. Beta=2.3, and Xi=30)
      (Nota: All images have been preliminary normalized to have the longuest side equals to 320 pixels)
      original images (longuest side = 320 pixels) ...................... tar.gz
      segmented images (pgm format) ............................................................... tar.gz | Results
      with contour super-imposed on the color image (ppm format) ........... tar.gz | Results
      Segmentation evaluation & comparison with other segmenters ... txt | png | Results
    • The source (C++//Linux) code of the SFSBM algorithm (only ppm image are supported) can be downloaded here tar
    • Berkeley image database http and Matlab code implementing the quantitative performance measures at: http

    PAGETM Algorithm

    thumbnail
    Non-local pairwise Gibbs energy based model for the HDR image compression problem
    Max Mignotte
    Journal of Electronic Imaging, 21(1):013016, anuary-March 2012.
      Abstract: In this work, a new Gibbs energy based compression model for the display of high-dynamic-range images. The proposed tone mapping method tends to exploit the biologically-inspired dynamic retina concept which is herein mathematically expressed via an image representation based on the specification of the statistical distributions of the non-local gradient magnitude. In this framework, the contrast reduction problem is expressed by a Gibbs energy with non-local pairwise pixel interactions whose cost function is minimized by a conjugate gradient descent procedure.
      ------------------
    • The source (in C++) code of our PAGETM algorithm and the makefile can be downloaded here: tar.gz
    • Compression results on some commonly used HDR images obtained by our algorithm: Results

    BCOCDM Algorithm

    thumbnail
    A bi-criteria optimization approach based dimensionality reduction model for the color display of hyperspectral images
    Max Mignotte
    IEEE Trans. on Geoscience and Remote Sensing,50 (2) : 501-513, Jan. 2012.
      Abstract: This paper proposes a new non-linear dimensionality reduction model based on a bi-criteria global optimization approach for the color display of hyperspectral images. The proposed fusion model is derived from two contradictory criteria of good visualization, namely; accuracy, with the preservation of spectral distance criterion and contrast, allowing the good separability of each observed existing material in the final visualized color image.
      ___________
    • The source (C++/Linux) code of the BCOCDM algorithm (only AVIRIS image are supported, can be downloaded here: tar.gz
    • Visualization results (ppm format) on the AVIRIS images obtained by our BCOCDM algorithm: Results
    • The hyperspectral data arre from the NASA JPL AVIRIS system and are kindly available at: http

    OBEHS Algorithm

    thumbnail
    An energy based model for the image edge histogram specification problem
    Max Mignotte
    IEEE Trans. on Image Processing, 21(1):379-386, Jan. 2012.
      Abstract: In this paper, we present a original energy-based model that achieves the edge histogram specification of a real input image. Experimental results are presented and several applications, of this model, such as a detail exaggeration procedure, a high-boost or enhancement filter, a normalization or a texture transfer technique, follow from this procedure.
      ------------------
    • The source (in C++) code of our OBEHS algorithm can be downloaded here ... tar.gz
    • We provide here image results obtained by our algorithm .................................... Results
    • The pseudo-code of our OBEHS algorithm can be downloaded here ................. pdf

    MD2S Algorithm

    thumbnail
    MDS-based multiresolution non-linear dimensionality reduction model for color image segmentation
    Max Mignotte
    IEEE Trans. on Neural Networks, 22(3): 447-460, March 2011.
      Abstract: In this work, we present an efficient coarse-to-fine multiresolution framework for multidimensional scaling (MDS) and demonstrate its performances to learn a nonlinear low-dimensional representation of the set of textural features of an image which can then subsequently and efficiently be exploited in a simple clustering based segmentation algorithm.
      ___________
    • Segmentation results on the Berkeley image database obtained by our algorithm MD2S (with parameters K=11 and xi=0.4)
      (Nota: All images have been preliminary normalized to have the longuest side equals to 320 pixels)
      original images (longuest side = 320 pixels) .......... tar.gz
      segmented images (pgm format) ..................................................... tar.gz | Results
      with contour super-imposed on the color image (ppm format) .. tar.gz | Results
      Performance measures of each segmented image ............................... txt | Results
      Comparison with other segmenters ................................... png
    • The source (C++) code of the MD2S algorithm (only ppm image are supported) can be downloaded here: tar
    • The Berkeley image database (with original image of size 481 x 321) is available at: http
    • The Matlab code mplementing the performance measures (used in this paper) was provided by A. Yang at: http

    SCKM Algorithm

    thumbnail
    A de-texturing and spatially-constrained K-means approach for image segmentation
    Max Mignotte
    Pattern Recognition Letters, 32(2):359-367, Jan. 2011.
      Abstract: We present a new and simple segmentation method only based on the k-means procedure and a two-step process. The first step relies on a original de-texturing procedure which aims at converting the input natural textured color image into a color image without texture, which will then be easier to segment. Once, this de-textured (color) image is estimated, a final segmentation is achieved by a K-means segmentation.
      ___________
    • Segmentation results on the Berkeley image database obtained by our algorithm SCKM (with parameters K1=9 and xi=0.37) (Nota: All images have been preliminary normalized to have the longuest side equals to 320 pixels)
      original images (longuest side = 320 pixels) ............ tar.gz
      segmented images (pgm format) ....................................................... tar.gz | Results
      with contour super-imposed on the color image (ppm format) .... tar.gz | Results
      Performance measures of each segmented image ................................. txt | Results
      Comparison with other segmenters ...................................... png
    • The source (C++/Linux) code of the SCKM algorithm (only ppm image are supported) can be downloaded here: tar
    • The Berkeley image database (with original image of size 481 x 321) is available at: http
    • The Matlab code implementing the performance measures (used in this paper) was provided by A. Yang at: http

    M4ICD Algorithm

    thumbnail
    A multiresolution Markovian fusion model for the color visualization of hyperspectral images
    Max Mignotte
    IEEE Trans. on Geoscience and Remote Sensing, 48(12):4236-4247, Dec. 2010.
      Abstract: We present a fusion/dimensionality reduction MRF model for the color display of hyperspectral images derived from the "preservation of spectral distance" criterion. In this framework, we also propose a computationally efficient coarse-to-fine conjugate gradient optimization method to minimize the cost function related to this energy-based fusion model
      ___________
    • The source (C++/Linux) code of the M4ICD algorithm, only AVIRIS image (with [SNR] are supported) can be downloaded here: tar.gz
    • Visualization results on the AVIRIS images obtained by our M4ICD algorithm and some AVIRIS images (in rfl format) : Results
    • The hyperspectral data are from the NASA JPL AVIRIS system and are kindly available at: http

    SFRBR Algorithm

    thumbnail
    Fusion of regularization terms for image restoration
    Max Mignotte
    Journal of Electronic Imaging, 19(3):333004-,July-Sept. 2010
      Abstract: We propose a restoration model associating a spatial and a frequential regularizers in order to better modeled the intrinsic properties of the original image to be recovered. An adaptive and rescaling scheme is proposed to balance the influence of these two different regularization constraints and allowing to prevent that a overwhelming importance for one of them prevail over the other in order to efficiently fuse them.
      ___________
    • The source (C++/Linux) code of the SFRBR algorithm (only squared pgm image are supported) can be downloaded here: tar
    • Restoration results on the cameraman and Lena image obtained by our algorithm SFRBR
      Original, degraded and restored images (pgm format): Results
      Comparison with other restoration (deconvolution) algorithms: png

    PRIF Algorithm

    thumbnail
    A label field fusion Bayesian model and its penalized maximum Rand estimator for image segmentation
    Max Mignotte
    IEEE Trans. on Image Processing,19(6):1610-1624, June 2010
      Abstract: This work proposes a new segmentation approach based on a MRF fusion model (in the maximum probabilistic rand index sense) which aims at combining several segmentation maps in order to finally get a more reliable and accurate segmentation result
      ___________
    • Segmentation results on the Berkeley database obtained by our algorithm PRIF (with parameters K1=18, K2=10 and K3=2) (Nota: All images have been preliminary normalized to have the longuest side equals to 320 pixels)
      Original images (longuest side = 320 pixels) ............................ tar.gz
      Segmented images (pgm format) ..................................................................... tar.gz | Results
      With contour super-imposed on the color image (ppm format) ................. tar.gz | Results
      Performance measures (PRI, VOI, GCE, BDE) of each segmented image ...... txt | Results
      Comparison with other segmenters .................................................. png
    • The Berkeley image database (with original image of size 481 x 321) is available at: http
    • The source (C++) code of the PRIF algorithm (only ppm image are supported) can be downloaded here: tar
    • The Matlab codes implementing the performance measures (used in this paper) is provided by A. Yang at: http

    NLREST Algorithm

    thumbnail
    A non-local regularization strategy for image deconvolution
    Max Mignotte
    Pattern Recognition Letters, 29(16):2206-2212, Dec. 2008
      Abstract: This paper proposes an Bayesian restoration (deconvolution) model exploiting a non parametric adaptive prior distribution derived from the non-local denoising model proposed by Buades et al
      ___________
    • The source code (C++/Linux) of the NLREST algorithm (only squared pgm image are supported to make the code easy to share) can be downloaded here tar
    • Restoration results obtained by our algorithm NLREST (see also the paper)
      Original, degraded and restored images (pgm format) ........................... Results
      Comparison with other restoration (deconvolution) algorithms .. png

    FCR Algorithm

    thumbnail
    Segmentation by fusion of histogram-based K-means clusters in different color spaces
    Max Mignotte
    IEEE Trans. on Image Processing,17(5):780-787, May 2008
      Abstract: This paper proposes a new and simple segmentation algorithm for natural textured images based on a K-mean based fusion procedure
      ___________
    • Segmentation results on the Berkeley database obtained by our algorithm FCR (with parameters K1=13, K2=6 & Kappa=0.135) (Nota: All images have been preliminary normalized to have the longuest side equals to 320 pixels)
      original images (longuest side = 320 pixels) .......................... tar.gz
      segmented images (pgm format) ..................................................................... tar.gz | Results
      with contour super-imposed on the color image (ppm format) .................. tar.gz | Results
      Performance measures (PRI, VOI, GCE, BDE) of each segmented image ....... txt | Results
      Comparison with other segmenters ...................................................... png
    • The Berkeley image database (with original image of size 481 x 321) is available at: http
    • The source code (C++/Linux) of the FCR algorithm (only ppm image are supported to make the code easy to share) can be downloaded here: tar
    • The Matlab code implementing the performance measures (used in this paper) is provided by A. Yang at: http

    DCT-EM Algorithm

    thumbnail
    DCT-based complexity regularization for EM tomographic reconstruction
    Max Mignotte, J. Meunier, J.-P. Soucy
    IEEE Trans. on Biomedical Engineering, 55(2):801-805, Feb. 2008
      Abstract: This paper introduces a simple algorithm for tomographic reconstruction based on the use of a DCT-based complexity regularization term
      ___________
    • The source code (C++/Linux) of the DCT-EM algorithm and the used synthetic phantom can be downloaded here: tar.gz

    DECWAV Algorithm

    thumbnail
    A post-processing deconvolution step for wavelet-based image denoising methods
    Max Mignotte
    IEEE Signal Processing Letters, 14(9):621-624, Sept. 2007
      Abstract: In this paper, we show that the performance of image denoising algorithms using certain wavelet transforms can be improved by a post-processing deconvolution step that takes into account the inherent blur function created by the considered wavelet based denoising system
      ___________
    • The source code (C++/Linux) of the DECWAV algorithm (only squared pgm image are supported to make the code easy to share) can be downloaded here: tar

      PS: Only the UDWT Deconv. algorithm is herein given; The curvelet code (to implement the CURV. Deconv. algorithm) can be easily incorporated in this source code by downloading the CurveLab toolbox implementing the (wrapping 2.0.1. version) Fast Discrete Curvelet Transform in C++ (written by L. Ying) and available on-line at: http://www.curvelet.org/

    SIMPAR Algorithm

    thumbnail
    Image denoising by averaging of piecewise constant simulations of image partitions
    Max Mignotte
    IEEE Trans. on Image Processing, 16(2):523-533, Feb. 2007
      Abstract: This paper introduces an image denoising algorithm which is based on an averaging process performed on a set of Markov Chain Monte-Carlo simulations of region partition maps constrained to be spatially piecewise uniform
      ___________
    • The source code (C++/Linux) of the SIMPAR algorithm and the test images can be downloaded here tar.gz
    • We provide here denoising results obtained by our algorithm: Results



    Licensing
    Copyright © 2011, Max Mignotte

    All rights reserved. All the softwares are distributed under the terms of a BSD-style license, i.e., redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
    1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
    2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
    3. Neither the name of the Université de Montréal nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.




    Last update: June 2011