Research Material

Multimodal Change Detection [NACCL] [M3CD] [FPMSMCD] [CICMMCD]
Fusion of label fields [GCEBFM] [FMBFM] [VOIBFM] [PRIF] [FCR]
Restoration/Deconvolution [SFRBR] [NLREST] [DECWAV]
Saliency Detection [PPMRF] [COSOV1NET] [MDSSME]
Symmetry/Asymmetry Detection [BISDA] [PAMGA]
Hyperspectral Compression [BCOCDM] [M4ICD]
Segmentation [MDSCCT] [MD2S][SCKM]
Reconstruction [SFSBM] [DCT-EM]
Sound Classification [ESCLBP]
HDR Compression [PAGETM]
Image Sonification [VFBAIS]
Contour Detection [BIEDA]
Image Browsing [SCVOI]
Vizualisation [OBEHS]
Denoising [SIMPAR]



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

  • MDSSME Algorithm

    thumbnail
    Fusion of color-based multi-dimensional scaling maps for saliency estimation
    M. Mignotte
    IntechOpen Book Chapter, Visual Object Tracking -Recent Advances, 2024
      Abstract: This paper presents an original energy-based model involving pixel pairwise constraints, in terms of color features, to which are then added constraints of higher levels of abstraction are exploited such as an over-segmentation and contour information, to the saliency map estimation problem.
      ___________
    • The source code of the MDSSME algorithm can be downloaded here [tar.gz]

    COSOV1NET Algorithm

    thumbnail
    CoSOV1Net: Cone-and Spatial-Opponency primary Visual cortex-inspired neural Network for lightweight salient object detection
    D. Ndayikengurukiye and M. Mignotte
    MDPI -Sensors, sect.: sensing & imaging, (spec. issue: object detect. based on vis. sens. & neur. net.), vol. 23 (14):6450, July 2023
      Abstract: This work presents CoSOV1Net, a novel lightweight (low FLOPS: 1.4 G and high FPS: 211.2 on GPU: Nvidia GeForce RTX 3090 Ti) salient object-detection neural network model, inspired by the cone- and spatial-opponent processes of the primary visual cortex (V1) which inextricably link color and shape in human color perception. Our proposed model is trained from scratch, without using backbones from image classification or other tasks and can be adapted to mobile environments and resource-constrained devices.
      ___________
    • The saliency map data of the CoSOV1Net algorithm on the ECSSD, DUT-OMRON, DUTS-TE, HKU-IS & THUR15K databases can be downloaded here [dir]

    PPMRF Algorithm

    thumbnail
    Saliency map estimation using a pixel-pairwise-based unsupervised Markov Random Field model
    M. Mignotte
    MDPI -Mathematics, 11(4):986, February 2023
      Abstract: This work presents an unsupervised Bayesian statistical approach, in a fully automatic pixel pairwise Markovian modeling framework,for the saliency map estimation problem.
      ___________
    • The source (C++/Linux) code and the data of the PPMRF algorithm (only ppm image are supported) can be downloaded here tar.gz

    NACCL Algorithm

    thumbnail
    MRF models based on a neighborhood adaptive class conditional likelihood for multimodal change detection
    M. Mignotte
    Journal, AI, Computer Science and Robotics Technology, vol. 8(4):110, April 2022
      Abstract: This work presents a new automatic Bayesian approach, relying on spatially adaptive class conditional likelihoods and adaptive to the considered imaging modality pair. Based on this likelihood model, defined for each pixel and each imaging modality, we compare the performance of different non-hierarchical or hierarchical Markovian estimators and priors on real multimodal satellite images.
      ___________
    • The source (C++/Linux) code and the data of the NACCL algorithm (only ppm image are supported) can be downloaded here tar.gz

    ESCLBP Algorithm

    thumbnail
    Environmental sound classification using local binary pattern and audio features collaboration
    O.K. Toffa and M. Mignotte
    IEEE Trans. on Multimedia, 23:3978-3985, November 2020
      Abstract: This paper presents a new approach to classify environmental sounds using a texture feature local binary pattern (LBP) and audio features collaboration.
      ___________
    • The source code of the ESCLBP algorithm can be downloaded here github

    CICMMCD Algorithm

    thumbnail
    Multimodal change detection using a circular invariant convolution model-based mapping
    R. Touati, M. Mignotte, M. Dahmane
    ASTES Journal, 5(5):1288-1298, October 2020
      Abstract: This work presents an original concentric circular invariant convolution model that aims at projecting the first satellite image into the imaging modality of the second image. This allows the two images to have identical statistics so that one can then effectively use a classical and monomodal change detection method based on a simple difference map.
      ___________
    • The source (C++/Linux) code and the data of the CICMMCD algorithm (only ppm image are supported) can be downloaded here tar.gz

    VFBAIS Algorithm

    thumbnail
    A hierarchical visual feature-based approach for image sonificationn
    O.K. Toffa and M. Mignotte
    IEEE Trans. on Multimedia, 23:706-715, April 2020
      Abstract: This work presents an image sonification system that strives to help visually impaired users access visual information via an audio (easily decodable) signal that is generated in real time when the users explore the image on a touch screen or with a pointer.

    • The source (C++/Linux) code of the VFBAIS algorithm can be downloaded here

    FPMSMCD Algorithm

    thumbnail
    A fractal projection and Markovian segmentation based approach for multimodal change detection
    M. Mignotte
    IEEE Trans. on Geoscience and Remote Sensing, 58(11):8046-8058, June 2020
      Abstract: This work presents a new parametric mapping strategy based on a modified geometric fractal decomposition and a contractive mapping approach allowing us to project the before image on any after imaging modality type. Once the projection is performed and that a pixelwise difference map between the two images (presented in the same imaging modality) is then binarized in the unsupervised Bayesian framework.
      ___________
    • The source (C++/Linux) code and the data of the FPMSMCD algorithm (only ppm image are supported) can be downloaded here tar

    M3CD Algorithm

    thumbnail
    Multimodal change detection in remote sensing images using an unsupervised pixel pairwise based Markov Random Field model
    R. Touati, M. Mignotte and M. Dahmane
    IEEE Trans. on Image Processing, 29(1):757-767, January 2020
      Abstract: This work presents an ubsupervised Bayesian statistical approach to the multimodal change detection (CD) problem in remote sensing imagery. More precisely, we use as observation field, a pixel pairwise modeling, whose distribution mixture are preliminary estimated. Once this estimation step is completed, the Maximum a posteriori (MAP) solution of the change detection map, based on the previously estimated parameters, is then computed with a stochastic optimization process.
      ___________
    • Results obtained by our algorithm Results
    • The source (C++/Linux) code and the data of the M3CD algorithm (only ppm image are supported) can be downloaded here tar

    GCEBFM Algorithm

    thumbnail
    A novel fusion approach based on the global consistency criterion to fusing multiple segmentations
    L. Khelifi and M. Mignotte
    IEEE Trans. on Systems, Man, and Cybernetics: Systems, 47(9):2489-2502, September 2017.
      Abstract: This paper presents a new fusion model which aims at combining multiple region-based segmentation maps to achieve a final improved segmentation result. The proposed fusion model is based on a consensus (energy) function derived from the perceptual and inherent multiscale nature of the Global Consistency Error which measures the extent to which one segmentation map can be viewed as a refinement of another segmentation.

    • Segmentation results on the Berkeley image database obtained by our algorithm
      (Nota: All images have been preliminary normalized to have the longuest side equals to 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 | pdf | Results
    • The source (C++/Linux) code of the GCEBFM algo (only ppm image are supported) can be downloaded here tar
    • Berkeley image database http and Matlab code implementing the quantitative performance measures at: http

    BIEDA Algorithm

    thumbnail
    A biologically-inspired framework for contour detection
    M. Mignotte
    Pattern Analysis and Applications, 20(2):365-381, May 2017.
      Abstract: This work presents a simple and efficient biological-inspired edge detection algorithm which attempts to model important structural features of the Human Visual System; such as the dynamic retina concept (i.e., the natural saccadic eye movement process which redirects the fovea's attention from one point of interest to another within the image) along with the inherent topological log-polar transformation of the retinal image into its cortical projection.

    • Segmentation results on the Berkeley image database obtained by our algorithm
      Contour-based Segmentation & performance measures: | Results | PR_Curve
    • The source (C++/Linux) code of the BIEDA algorithm (only ppm image are supported) can be downloaded here tar.gz
    • Berkeley image database http

    SCVOI Algorithm

    thumbnail
    Segmentation data visualizing and clustering
    A. Khlif and M. Mignotte
    Multimedia Tools and Applications, 76(1):1531-1552, January 2017.
      Abstract: We propose here a new framework that helps the user to browse through large image database in an intuitive and efficient manner based on their segmentation-based descriptive content or the geometrical layout and shapes of the different objects segmented in the scene.

      Results obtained by our algorithm:
    • Increased resolution of certain images presented in the paper: Figures
    • Mapping & Clustering results on the Berkeley image database (BSDS300) .. html
    • Mapping results on the Weizmann image database (Weizmann200) : html
    • Mapping results on the Stanford image database (Stanford715) : html
    • Mapping results on the Msrc image database (Msrc590) : html

    The source (C++/Linux) code of the SCVOI algorithm (only ppm image are supported) can be downloaded here tar.gz

    BISDA Algorithm

    thumbnail
    Symmetry detection based on multiscale pairwise texture boundary segment interactions
    M. Mignotte
    Pattern Recognition Letters, 74:53-60, Elsevier, February 2016.
      Abstract: In this work, we propose a new unsupervised and simple approach to local symmetry detection in natural images. The proposed model consists in quantifying the presence of a partial medial axis segment, existing between each pair of (preliminary detected) line segments delineating the boundary of two textured regions, by a set of heuristics related both to the geometrical structure of each pair of line segment and its ability to locally delimit a homogeneous texture region in the image.

    • Symmetry detection results on the Berkeley image database (BSDS300) obtained by our BISDA algorithm:
      Symmetry detection (pgm format) : tar.gz
      F measure (curve, global and for each image) : txt | html
      (Symmetry detection obtained only in Cartesian html in log-polar space html)
    • The source (C++/Linux) code of the BISDA algorithm (only ppm image are supported) can be downloaded here tar
    • Local Symmetry for the Berkeley image database and Matlab code implementing the quantitative performance measures at: http

    PAMGA Algorithm

    thumbnail
    A perceptual maps for gait symmetry quantification and pathology detection
    A. Moevus, M. Mignotte, J. de Guise and J. Meunier
    BioMedical Engineering OnLine, 14(1):99, October 2015.
      Abstract: A new method to estimate a perceptual color map of asymmetries from a video acquired by a depth sensor (Kinect) of a patient walking on a treadmill is introduced. The proposed gait analysis method relies on the fact that the healthy walk has (temporally shift-invariant) symmetry properties in the coronal plane. A perceptual color map of asymmetries is the compression of a patient's video mapped into a single color image in such manner that asymmetries of the body movements in the human gait cycle are visible and quantifiable for pathology detection, follow up or (early) diagnosis of many different types (neurological, muscular, orthopedic, etc.) of diseases.

    • The source (C++/Linux) code of the PAMGA algorithm can be downloaded here tar.gz
    • The database of depth map video sequences can be downloaded here tar.gz
    • The C function to read a depth map video sequence can be downloaded here txt

    FMBFM Algorithm

    thumbnail
    A precision-recall criterion based consensus model for fusing multiple segmentations
    C. Hélou and M. Mignotte
    International Journal of Signal Procesing, Image Processing and Pattern Recognition, 7(3): 61-82, July 2014.
      Abstract: This work presents a new fusion model derived from the well-known precision-recall criterion for combining multiple quickly estimated segmentation maps of the same scene to achieve a final improved segmentation.
      ___________
    • Segmentation results on the Berkeley image database obtained by our algorithm
      (Nota: All images have been preliminary normalized to have the longuest side equals to 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 | pdf | Results
      Contour-based performance measures: | Results
    • The source (C++/Linux) code of the FMBFM algorithm (only ppm image are supported) can be downloaded here tar.gz
    • Berkeley image database http and Matlab code implementing the quantitative performance measures at: http

    VOIBFM Algorithm

    thumbnail
    A label field fusion model with a variation of information estimator for image segmentation
    M. Mignotte
    Information Fusion, Elsevier, vol. 20: 7-20, January 2014.
      Abstract: This work presents a new fusion model of segmentations based on a consensus (energy) function derived from the recent information Theory based variation of information criterion, proposed by Meila.

    • Segmentation results on the Berkeley DataSet image database (BSDS300) obtained by our algorithm
      (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 | pdf | Results
      F-Measure on the BSDS300: Results
    • Segmentations on the Weizmann DataSet
      Score: 1obj/2obj/segmentations: [1obj] [2obj] [gz]
      VOIBFM with 4 classes at maximum: [1obj] [2obj] [gz]
    • The source (C++//Linux) code of the VOIBFM algorithm (only ppm image are supported) can be downloaded here tar.gz
    • Berkeley image database http and Matlab code implementing the quantitative performance measures at: http

    SFSBM Algorithm

    thumbnail
    A non-stationary MRF model for image segmentation from a soft boundary map
    M. Mignotte
    Pattern Analysis & Aplications, 17(1)-129-139, 2014.
      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 SFSBM (with param. Beta=2.3, and Xi=30)
      (Nota: All images have been preliminary normalized to have the longuest side equals to 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.gz
    • Berkeley image database http and Matlab code implementing the quantitative performance measures at: http

    MDSCCT Algorithm

    thumbnail
    MDS-based segmentation model for the fusion of contour and texture cues in natural images
    M. Mignotte
    Computer Vision and Image Understanding, 116(9):981-990, Sept. 2012.
      Abstract: In this work, we propose an original segmentation model based on a preliminary spatially adaptive non-linear dimensionality data reduction step integrating contour and texture cues which aims at converting an input texture image into a noisy color image in order to greatly simplify its segmentation with a simple K-mean clustering scheme.

    • Segmentation results on the Berkeley image database obtained by our algorithm MDSCCT (Nota: All images have been preliminary normalized to have the longuest side equals to 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-based performance measures of each segmented image: txt | Results
      Contour-based performance measures: | Results
    • The source (C++) code of the MDSCCT algorithm (only ppm image are supported) can be downloaded here tar.gz
    • 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
    M. Mignotte
    Journal of Electronic Imaging, 21(1):013016, January-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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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
    M. 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