(last update : 03/02/99)
Jean-Michel Jolion and Azriel Rosenfeld
Category: Book
Abstract: This book present a comprehensive survey of the pyramid framework for visual tasks related to early vision such as object detection and delineation, texture analysis, contour extraction, image processing (contrast enhancement and smoothing), motion analysis, ... We first introduce a general presentation of the pyramid model with references to hardware architectures. Then, the main part of the book is devoted to image segmentation. Extensions such as depth analysis, data-flow pyramid are proposed in the last part of the book.
Publication-Status: Copyright KLUWER Pub. Comp., 1994
Keywords: Pyramid, Early viion, Computer Vision, Segmentation, Motion, Multiresolution, Hierarchical processing, Motion, Texture, Contour
Size: 222p
Souheil Ben Yacoub and Jean-Michel Jolion
Category: Journal
Abstract: A hierarchical line and segment extraction algorithm, based on a pyramid, is described. We first detect lines in small windows using a Hough transform and then merge these lines using distanbce criteria thus avoiding a re-accumulation process at each level of the pyramid structure. The hierarchical merging process is performed on lines and not segments, which is obviously more efficient since there are much moore segments than lines. The spilitting algorithm of a line in segments is performed at the top of the pyramid. The proposed approach is compared to other related works based on hierarchical feature extraction. We show that our approach combines the advantages of other works and avoids the backdraws (quantization effect, lack of robustness).
Publication-Status: IEE Vision, Image and Signal Processing, 142(1), 1995, 7-14.
Keywords: Hough transform, pyramid, line/segment extraction.
Souheil Ben Yacoub and Jean-Michel Jolion
Category: Journal
Abstract: We propose to study the relationship between the hierarchical Hough transform and classical algorithms for partioning curves into segments. We show that the set of segments found by classical algorithms is a subset of the segments found by the hierarchical Hough transform. This new relationship allows to have a better understanding of the behaviour of some parameters of the hierarchical Hough transform.
Publication-Status: Pattern Recognition Letters, 16(4), 1995, 389-398.
Keywords: Hough transform, pyramid, line/segment extraction, polygonal curve approximation.
Antoine Manzanera and Jean-Michel Jolion
Category: Journal
Abstract: We present a new class of hierarchical structures known as the irregular pyramid. This pyramid is characterized by a non regular subsampling defined as a fiunction of the spatial location of the points. This hierarchical structure is made as consistent as possible with the human visual mapping. This novel structure allows both fine sampling in the focus area and coarse sampling elsewhere in the scene so resulting in smaller images. Examples are shown as well as an exploratory vision application in motion detection.
Publication-Status: Traitement du Signal, 1995.
Keywords: Pyramid, multiresolution, motion.
Jean-Michel Jolion
Category: Journal
Abstract: We propose a new approach of contrast in digital image using a multiresolution framework. Contrast enhencement is the main application by mean of an iterative process. Moreover, we show that when iterated, this process ends up with a simplified, e.g., binary, image from any initial image. Several examples are presented all over the paper showing the performance of our algorithm on synthetic as well as real scenes.
Publication-Status: Traitement du Signal, 11(3), 1994, 245-255.
Keywords: Pyramid, multiresolution analysis, contrast, binary images.
Ernest Chiarello, Jean-Michel Jolion and Claude Amoros
Category: Journal
Abstract: In the framework of landscape ecology, we propose to modelize the spatial organization of vegetation in river floodplains. This model is based on the generation of non-homogeneous anisotropic random patterns defined on a sqaure lattice given the knowledge of the ecological functioning of these ecosystems. This model proceeds in two steps.
First, geographic information is computer (distances to the nearest dead arm and to the nearest active channel, convexity and concavity, ...) using the information about whole hydrographic network of channels (presently and pastly created by the fluvial dynamic). This leads to a first tesselation of the image. Then a set of probabilities of occurence of each vegetation type is assigned to each site.
Second a random number, weighted by the probability of the observed vegetation, is assigned to each site; then the consolidation can process. The weakest sites are eliminated while the strongest, called the survivors, extend their frontiers and incorporate the sites just deleted. The two processes, decimation and expansion, are recursive and parallel.
At every hierarchical level, islands and channels contents are updated; this allows or not the expansion of vegetation, with respect to the propabilities. Each vegetation of each element has its own expansion criteria (vegetation, geographical proximity or random choice).
The resulting patterns are not segmentations of the tesselation defined by initial geographic information. On the contrary, they present a stochastic appearance which reflects both the sqtochastic appearance and the spatial determinism of vegetation spatial organization in river floodplains due to the fluvial dynamic.
Publication-Status: Acta Stereologica, 12(2), 1993, 255-261.
Keywords: Stochastic pyramid, spatial pattern, modelization, river floodplain.
Azriel Rosenfeld and Jean-Michel Jolion
Category: Journal
Abstract: When a local operation is performed on the pixels in an array, the new value of the pixel is a function of the old values of the pixel and its neighbors. This paper introduces the more general concept of local operations on labelled dot patterns, where the new label of a dot is a function of the old labels of the dot and a set of its neighbors (e.g. its Voronoi neighbors) Such operations may change the positions of the dots, in addi- tion to changing their 'values'. We illustrate these ideas by giving examples of operations that perform local feature detection (e.g. isolated dot detection, cluster edge detection, dotted curve detection) and 'enhancement' (e.g. 'smoothing' the dot spacing or 'sharpening' the edges of diffuse clusters), as well as 'morphological' operations. Local operations on label- led graphs are also briefly discussed.
Publication-Status: Pattern Recognition Letters, 9, 1989, 225-232.
Keywords: Dots Patterns, Voronoi Diagram
Jean-Michel Jolion and Azriel Rosenfeld
Category: Journal
Abstract: If a feature space contains a set of clusters and background noise, it may be difficult to extract the clusters correctly. In particular, when we use a partitioning scheme such as k-means clustering, where k is the correct number of clusters, the background noise points are forced to join the clusters, thus biaising their statistics. This paper describes a pre-processing technique that gives each data point a weight related to the density of data points in its vicinity. Points belonging to clusters thus get relatively high weights, while background noise points get relatively low weights. k-means clustering of the resulting weighted points converges faster and yields more accurate clusters.
Publication-Status: Pattern Recognition, 22(5), 1989, 603-607
Keywords: Cluster detection, Dot density measurement, Dot patterns, k-means clustering, Noise clustering
Jean-Michel Jolion and Azriel Rosenfeld
Category: Journal
Abstract: This paper describes a divide-and-conquer Hough transform technique for detecting a given number of straight edge or lines in an image. This technique is designed for implementation on a pyramid of processors, and requires only O(log n) computational steps for an iage of size n x n.
Publication-Status: Pattern Recognition Letters, 9, 1989, 343-349
Keywords: Pyramid, Hough Transform, Line detection, Divide-and-conquer
Jean-Michel Jolion and Azriel Rosenfeld
Category: Journal
Abstract: The bimodality of a population P can be measured by dividing its range into two intervals so as to maximize the Fisher distance between the resulting two subpopulations P1 and P2. If P is a mixture of two (approximately) Gaussian subpopulations, then P1 and P2 are good approximations of the original Gaussian, if their Fisher distance is great enough. For a histogram having n bins this method of bimodality analysis requires n-1 Fisher distance computations, since the range can be divided into two intervals in n-1 ways. The method can also be applied to 'circu- lar' histograms, e.g. of populations of slope or hue values; but for such histograms it is much more computationaly costly, since a circular histogram having n bins can be devided into two intervals (arcs) in n(n-1)/2 ways. The cost can be reduced by performing bimodality analysis on a 'reduced-resolution' histogram having n/k bins; finding the subdivision of this histogram that maximizes the Fisher distance; and then finding a maximum Fisher distance subdivision of the full-resolution histogram in the neighborhood of this subdivision. This reduces the required number of Fisher distance computations to n(n-1)/(2k^2) + O(k). For histograms representing mixtures of two Gaussians, this method was found to work well for n/k as small as 8.
Publication-Status: Pattern Recognition Letters, 10, 1989, 201-207
Keywords: Pyramid, Histogram thresholding, Fisher distance, Multiresolution, Bimodality analysis
Jean-Michel Jolion, Peter Meer and Azriel Rosenfeld
Category: Journal
Abstract: A new method for delineation of compact objects in image pyramids is presented. The borders of the objects are detected in a low-resolution representation of the input, a higher level of the pyramid. The pixels on the two sides of an edge are the roots for two classes (object and background). The two classes are employed in two independent top-down tree growing processes. The information is passed downward by adjusting confidence measures. The employment of multiple roots defined on the smoothed representation of the input contributes to the robustness of the method at very low signal-to-noise ratios.
Publication-Status: Pattern Recognition Letters, 11, 1990, 107-115
Keywords: Image pyramids, multiresolution methods, feature extraction, noise smoothing
Peter Meer, Jean-Michel Jolion and Azriel Rosenfeld
Category: Journal
Abstract: The only information available to a blind noise variance estimation algorithm is the corrupted image and the white nature of the zero mean Gaussian noise. The proposed algorithm recovers the variance of the noise in two steps. First, the sample variances are computed for square cells tessellating the noisy image. Several tessellations are applied with the size of the cells increasing fourfold for consecutive tessellations. The four smallest sample variance values (the outcomes of the first four order statistics) are retained for each tessellation and combined through an outlier analysis into one estimate. The different tessellations thus yield a variance estimate sequence. In the second part of the algorithm, the value of the noise variance is determined from this variance estimate sequence. We have applied the blind noise variance algorithm to 500 noisy 256 x 256 images derived from seven prototypes of classes often employed in computer vision and image processing. In 98 percent of the cases the relative estimation error was less than 0.2 with an average error of 0.06. Application of the algorithm to differently sized images is also discussed. All the operations in the algorithm are parallel and if they are implemented on an image pyramid, the variance of the noise is recovered in O[log(image_size)] processing time.
Publication-Status: IEEE Trans. on Pattern Analysis and Machine Intelligence, 12, 1990, 216-223
Keywords: Image pyramids, Noise estimation, Order statistics
Jean-Michel Jolion, Peter Meer and Samira Bataouche
Category: Journal
Abstract: A novel clustering algorithm based on the minimum volume ellipsoid (MVE) robust estimator recently introduced in statistics is proposed. The MVE estimator identifies the least volume region containing h percent of the data points. The clustering algorithm iteratively partitions the space into clusters without a priori information about their number. At each iteration, the MVE estimator is applied several times with values of h decreasing from 0.5. A cluster is hypothesised for each ellipsoid. The shapes of these clusters are compared with shapes cooresponding to a known unimodal distribution by the Kolmogorov-Smirnov test. The best fitting cluster is then removed from the space, and a new iteration starts. Constrained random sampling keeps tha amount of computation low. The clustering algorithm was successfully applied to several computer vision problems formulated in the feature space paradigm: multithresholding of gray level images, analysis of the Hough space, range image segmentation.
Publication-Status: IEEE Trans. on Pattern Analysis and Machine Intelligence, 13, 1991, 791-802
Keywords: Clustering, Feature space, Hough transform, Multithresholding, Range image segmentation, Robust estimation
Jean-Michel Jolion and Annick Montanvert
Category: Journal
Abstract: The adaptive pyramid is a new framework for 2D image analysis. It is based on the principle of local evidence accumulation for global interpretation. The adaptive structure of this pyramid is a new approach of reduced resolution. We introduce the concept of surviving cell as a local maximum of an interest operator. A root is then defined as a particular surviving cell. Examples are shown in segmentation for different images. The limits of this technique are presented on an example of a highly textured image. A particular top-down process is presented in the context of shape decomposition.
Publication-Status: CVGIP: Image Understanding, 55, 1992, 339-348
Keywords: Pyramid, Image segmentation
Muriel Mattenet and Jean-Michel Jolion
Category: Journal
Abstract: This paper presents the finger-print of a contour technique. It is a new representation allowing a fast and robust extraction of structural differences and resemblances between shapes. The examples shown are a part of a comparison of contour drawings of France.
Publication-Status: Mappemonde, 3, 1992, 5-9
Keywords: Handwritting analysis, Fingerprint, Scale-space
Jean-Michel Jolion
Category: Journal
Abstract: In this paper, we address the problem of methodologies for computer vision. In the first part we will present a brief survey of the Marr paradigm, e.g., what David Marr called his philosophy. We will emphasize the sequence of hypothesis which progressively makes the scene recovery approach explicit as well as the limitations of this approach. We then present the goal-directed approach as an alternative to the recovery school: behaviorism versus reconstructionism. We show that this dichotomy is not the only possible one and introduce the idealism versus empiricism dichotomy. We propose some directions toward a new methodology in a systemic framework involving another, higher-level, methodological dichotomy: systemism versus reductionism. In this new framework we try to exploit of all the sources of constraints, and, thereby, to reconcile some of the previous approaches like recovery school and purposive vision.
Publication-Status: CVGIP: Image Understanding, 59(1), 1994, 53-71.
Keywords: Marr's three levels, methodologies, Computer vision, Scene recovery, Active vision, Smart sensing, Systemic, Purposive vision
Jean-Michel Jolion
Category: Journal
Abstract: The problem of the delineation of a set of geometric features in a hierarchical environment is addressed. A brief summary of a feature extraction and bottom-up propagation technique is given. A top-down process is proposed toward feature delineation based on a general model for hierarchical dependencies. The main idea is to make explicit the compromise between local and global consistency in the context of hierarchical reasoning by means of consecutive levels of interaction. Several examples in multi-thresholding of gray images, line detection and delineation, and range image segmentation are presented.
Publication-Status: Pattern Recognition, 26(9), 1993,1295-1304.
Keywords: Hierarchical reasoning, Range image segmentation, Gray image multithresholding, Line detection and delineation, Geometric feature
Jean-Michel Jolion
Category: Journal
Abstract: L'analyse d'images numeriques est un domaine ou les notions de parallelisme et de recursivite sont frequeemment utilisees. L'architecture pyramidale permet de mettre efficacement en oeuvre ces deux notions. Dans cet article, nous presentons trois aspects des pyramides prises tour a tour comme un modele architectural pour une machine massivement parallele, un modele pour la representation de donnees hierarchisees et enfin un modele algorithmique pour les processus d'extraction rapide d'information.
Publication-Status: Traitement du Signal, 7(1), 1990, 5-17.
Keywords: Analyse d'images, Pyramide, Modelisation, Parallelisme, Recursivite, Architecture
Jean-Michel Jolion and Patrick Prevot
Category: Journal
Abstract: La binarisation est une etape classique du traitement d'une image. Nous presentons dans cet article une procedure rapide et performante pour des images obtenues en microscopie electronique. Cette methode se caracterise par la prise en compte de l'intensite lumineuse ainsi que de la morphologie de l'image.
Publication-Status: Traitement du Signal, 3(3), 1986, 153-158.
Keywords: Image binaire, microscopie electronique, decomposition d'histogramme
Jean-Michel Jolion, Peter Meer and Azriel Rosenfeld
Category: Conference
Abstract: A novel clustering algorithm based on the minimum volume ellipsoid (MVE) robust estimator recently introduced in statistics is proposed. The MVE estimator identifies the least volume region containing h percent of the data points. The algorithm iteratively partitions the space into clusters without a priori information about their number. At each iteration, the MVE estimator is applied several times with values of h decreasing from 0.5. A cluster is hypothesised for each ellipsoid. The shapes of these clusters are compared with shapes cooresponding to a known unimodal distribution by the Kolmogorov-Smirnov test. The best fitting cluster is then removed from the space, and a new iteration starts. Constrained random sampling keeps tha amount of computation low. The clustering algorithm was successfully applied to several computer vision problems formulated in the feature space paradigm: decomposition of gray level image histograms and analysis of the Hough space.
Publication-Status: First Int. Conf. on Robust Computer Vision, Seattle, WA, USA, october 1-3, 1990, 339-351
Keywords: Robust statistic, Classification, Image segmentation
Samira Bataouche and Jean-Michel Jolion
Category: Book chapter
Abstract: We introduce a robust statistic estimator in a pyramidal environment in order to discriminate different image properties. After a presentation of the initial algorithm, we will see the fundamental points we had to take into account to implement it in a hierarchical processing environment. Examples of different soace dimension applications are given, like the detection of lines in grey scale images and the detection of planes in range images.
Publication-Status: Progress in Image Analysis and processing II, Cantoni V., Ferretti M., Levialdi S., Negrini R. and Stefanelli R. eds., World Scientific, Singapore, 1992, 510-517
Keywords: Pyramid, Image segmentation, Geometric feature delineation
Jean-Michel Jolion
Category: Conference
Abstract: We present a novel technique toward contrast enhancement using a multiresolution representation of an image. The contrast pyramid is extracted from this representation. The method enhances the image by means of modification of the contrast values at multiple scales, i.e. in the multiresolution domain. The enhanced image is then recovered by reversing the steps used in the construction of the multiresolution representation.
Publication-Status: Proc. Int. Conf. Image Processing: Theory and Applications, San Remo, Italy, Vernazza G., Venetsanopoulos A.N. and Braccini C., eds., Elsevier, Amsterdam, 1993, 205-208
Keywords: Pyramid, Multiresolution, Contrast enhancement, Thresholding
Samira Bataouche and Jean-Michel Jolion
Category: Conference
Abstract: In this paper, we study the hierarchical reasoning and especially the pyramidal processing of image data. First of all, we are concerned with the way information will be transfered from one pyramidal cell to another when thses cells work asynchronously. As example, we will consider the asynchronous hierarchical extraction of connected components in binary images.
Publication-Status: Proc. Int. Conf. Image Processing: Theory and Applications, San Remo, Italy, Vernazza G., Venetsanopoulos A.N. and Braccini C., eds., Elsevier, Amsterdam, 1993, 267-270.
Keywords: Pyramid, Connected component
Jean-Michel Jolion et Annick Montanvert
Category: Conference
Abstract: Nous presentons dans cet article un principe de construction d'une pyramide logicielle, baptisee la pyramide adaptative. Cette structure est appropriee a la detection et a la localisation de regions dans une scene bidimensionelle. La construction n'est pas rigide comme pour une structure pyramidale classique, mais tient compte du contenu du niveau courant de la pyramide. Ceci est realise par appel a un operateur d'interet, evalue en chaque point du niveau courant, pour guider la selection des points suivants. Aux differents niveaux vont ressortir des points contrastes avec leur voisinage, traduisant la presence d'objets dans l'image. La delimitation de ces objets est obtenue progressivement dans la phase de descente, en faisant intervenir des contraintes de forme.
Publication-Status: Proc. 7eme Congres AFCET/INRIA Reconnaissance des Formes et Intelligence Artificielle, Paris, Novembre 1989, 197-206.
Keywords: Pyramid, Multiresolution, Image segmentation
Jean-Michel Jolion, Peter Meer and Azriel Rosenfeld
Category: Conference
Abstract: Nous presentons une nouvelle approche de certains problemes de vision artificielle grace a l'estimateur de l'ellipsoide de volume minimal applique a l'espace des caracteristiques. Les classes sont mises en evidence par une approche multiresolution. Cet algorithme a ete applique avec succes a des problemes tres divers tel que le multi-seuillage d'images noir et blanc, ou couleur, la recherche de segments de droites (analyse de l'espace de Hough), la recherche de surfaces planes ou biquadratiques dans des images de profondeur.
Publication-Status: Proc. XXIIeme Journees de Statistique, Tours, France, 28 mai - 1er juin, 1990, 70-72.
Keywords: Robust statistic, Clustering, Image segmentation
Jean-Michel Jolion
Category: Conference
Abstract: In this paper, we propose an application of detection and localization of motion using both multiresolution and robust statistics. We show that simples hypotheses that are usually rejected could be used thanks to redundant information addition (multiresolution) and efficient statistic tools such as M-estimators. Many examples are presented on synthetic and real sequences of images.
Publication-Status: Proc. 8eme Congres AFCET/INRIA Reconnaissance des Formes et Intelligence Artificielle, Lyon, Novembre 1991, 557-566.
Keywords: Pyramid, Multiresolution, Motion analysis
Samira Bataouche and Jean-Michel Jolion
Category: Conference
Abstract: In this paper, we present the use of a robust statistical method of clustering in a hierarchical environment. So, after a presentation of the robust clustering algorithm, we will see the fundamental points we had to take into account to implement it in a pyramidal processing environment. Examples of applications are given, like the thresholding of gray scale images or the detection of planes in range images.
Publication-Status: Proc. 8eme Congres AFCET/INRIA Reconnaissance des Formes et Intelligence Artificielle, Lyon, Novembre 1991, 549-555.
Keywords: Pyramid, Hierarchical clustering, Image segmentation
Thierry Excoffier and Jean-Michel Jolion
Category: Conference
Abstract: La technique de fusion d'images par fusion de leur decomposition frequentielle est presentee puis generalisee avec la prise en compte d'une composante temporelle. La nouvelle pyramide Laplacienne ainsi obtenue permet une separation des differents mouvements et autorise une recombinaison plus riche. Des exemples sont proposes en creation de fondu enchaine et en fusion de sequences d'images d'objets en mouvement. Nous concluons sur une discussion des potentialites et developpements eventuels de la methode.
Publication-Status: Proc. 11eme Conf. MICAD, Paris, 11-14 fevrier 1992, 219-229.
Keywords: Pyramid, Multiresolution, Motion, Image synthesis
Jean-Michel Jolion
Category: Conference
Abstract: In this paper, we address the problem of methodologies for the design of computer vision systems. In the first part we will present a brief survey of the Marr paradigm, e.g., what David Marr called his philosophy. We will emphasize the sequence of hypothesis which progressively makes the scene recovery approach explicit as well as the limitations of this approach. We then present the goal-directed approach as an alternative to the recovery school: behaviorism versus reconstructionism. We show that this dichotomy is not the only possible one and introduce the idealism versus empiricism dichotomy. Finally, we propose a global constraint analysis.
Publication-Status: Proc. Int. Summer School on Real-Time Systems Computer Vision and Process Automation, Tempus Project 0962, 13-20 juin 1992, Krakow, Poland.
Keywords: Computer Vision methodologies, Computer Vision system
Jean-Michel Jolion
Category: Conference
Abstract: Beaucoup d'applications en vision par ordinateur font appel aux techniques statistiques et tout particulierement l'estimation et la classification. Cependant, en vision par ordinateur, le bruit est un concept incontournable des que l'on se conforte a des donnees reelles. Pour tenter de repondre a cet inconvenient, nous introduisons les concepts de statistiques robustes et leurs utilisations recentes en vision par ordinateur.
Publication-Status: Proc. 1er Colloque Africain sur la Recherche en Informatique, Yaounde, Cameroun, 14-20 octobre 1992.
Keywords: Robust Statistic, Estimation, Image segmentation
Ernest Chiarello, Jean-Michel Jolion and Claude Amoros
Category: Conference
Abstract: Nous proposons un modele d'organisation spatiale des plaines alluviales, autour de la pyramide stochastique. Mais au lieu de partir d'une image aleatoire, et de regrouper les regions selon des criteres eux aussi aleatoires, nous proposons de construire une image initiale coherente. Ainsi, en tenant compte de probabilites a priori calculees a partir d'une image de contraintes, les pixels sont regroupes en fonction de leurs positions geographiques dans cette image. Pour chacun, la probabilite d'observer tel type de vegetation plutot qu'un autre depend du stade d'evolution, de sa distance a l'amont ou a l'aval, de la proximite de courbures fortes, de sa distance a un chenal actif. Ensuite, le processus de decimation-expansion de la pyramide les regroupe en des formes particulieres qui respectent les criteres d'organisation dans l'espace.
Publication-Status: Proc. Colloque Geometrie Discrete en Imagerie: Fondements et Applications, Grenoble, 15-18 septembre 1992, 255-269.
Keywords: Image synthesis, Stochastic pyramid, Adaptive pyramid
Antoine Manzanera and Jean-Michel Jolion
Category: Conference
Abstract: Nous introduisons une nouvelle classe de structures hierarchiques irregulieres ou la nature du sous-echantillonnage est une fonction de la position spatiale. Cette structure s'apparente au mapping retinien (notion de fovea et de peripherie). Ce nouveau mecanisme permet de conserver un echantillonnage precis dans la zone de focus tout en resumant le restant de la scene. Nous montrons que cette structure peut etre facilement obtenue sans faire appel a des mecanismes d'approximation d'une theorie continue.
Publication-Status: Proc. Colloque Geometrie Discrete en Imagerie: Fondements et Applications, Strasbourg, 20-21 septembre 1993, 98-107.
Keywords: Pyramid, Multiresolution, Irregular tessellation
Ernest Chiarello, Jean-Michel Jolion and Claude Amoros
Category: Journal
Abstract: The stochastic pyramid is introduced in landscape ecology in order to propose a spatial organization model of ecological units in river floodplains. It consists in articulating this parallel generator of random patterns with ecological processes. The stochastic pyramid, based on a hierarchy of seed structures, is used to make regions grow in order to obtain random patterns relating to an a priori model. The first level, represented by the lattice, is obtained after a biased random draw and the upper levels are built with a double controlled process of decimation and expansion of the seeds. The resulting pictures present both deterministic and stochastic spatial patterns according to ecological reality.
Publication-Status: Pattern Recognition, 29(1), 1996, 61-75.
Keywords: Stochastic Pyramid, Pattern generation, Irregular tessellation, Ecological model
Ernest Chiarello, Claude Amoros, Guy Patou and Jean-Michel Jolion
Category: Journal
Abstract: The stochastic pyramid is introduced in landscape ecology in order to propose a spatial organization model of ecological units in river floodplains. It consists in articulating this parallel generator of random patterns with ecological processes. The stochastic pyramid, based on a hierarchy of seed structures, is used to make regions grow in order to obtain random patterns relating to an a priori model. The first level, represented by the lattice, is obtained after a biased random draw and the upper levels are built with a double controlled process of decimation and expansion of the seeds. The resulting pictures present both deterministic and stochastic spatial patterns according to ecological reality.
Publication-Status: Landscape Ecology, 1995.
Keywords: Stochastic Pyramid, Pattern generation, Irregular tessellation, Ecological model
Jean-Michel Jolion
Category: Conference
Abstract: We propose in this paper some trends toward a new approach of artificial vision. We first introduce the main points related to already proposed methodologies based on the Marr's paradigm. We then show that the inside behavior of the observer is a source of information as well as the physical properties of visibles surfaces. Examples are proposed in the particular case of connected components extraction in a binary image using a data flow pyramid computer showing that "how" we process information may be more important than "what" we actually process.
Publication-Status: 8th Int. Workshop on Theoretical Foundation of Computer Vision, March 1996.
Keywords: Pyramid, Behavioral model, Temporal analysis
Michel Borowy and Jean-Michel Jolion
Category: Conference
Abstract: We present in this paper some new trends toward a data flow pyramid model. This framework is based on classic hierarchical graph construction algorithm. The basic idea underlying our work is to remove the constraint between successive levels construction. We map this hierarchical structure in a rigid pyramid in order to better control this new process, and more particularly focus our work on a data flow communication management problem. We show how this model can be used for fast detection and delineation of image feature.
Publication-Status: 4th Int. Workshop on Parallel Image Analysis, December 1995, 193-202.
Keywords: Pyramid, Blob detection and delineation, Temporal analysis
Christophe Duperthuy and Jean-Michel Jolion
Category: Conference
Abstract: The representation of visual data extracted from pictures is of importance according to the performances of the visual system to conceive. Such a representation should explicit all the real physical attributes, while being robust. In this way, we propose a new Primal Sketch based on simplified pyramidal representation of the scene to deal with: it results in robustness and fast calculations (using a massive parallelism), especially with the use of 3 x 3 binary masks. We obtain then a set of attributes cards, such as noises, textures shapes and blobs, for various resolution levels.
Publication-Status: 4th Int. Workshop on Parallel Image Analysis, December 1995, 53-64.
Keywords: Pyramid, Primal sketch, Contrast image
Jean-Michel Jolion
Category: Report
Abstract: In this paper, we introduce a new scheme for image simplification which enables fast pre-processing of image data and which is very useful for edge detection as it reduces the ambiguity related to the choice of the parameters characterizing these edge detectors. Mainly, this process makes use of a new approach of contrast in a multiresolution framework and more particularly in pyramids. Moreover, this process is a very simple way to efficiently combine multiresolution information in a single image.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 02.96
Keywords: Image simplification, contrast enhancement, multiresolution, early vision, pyramid, edge detection
Jean-Michel Jolion
Category: Report
Abstract: We propose in this paper some trends toward a new approach of artificial vision. We first introduce the main points related to already proposed methodologies based on the Marr's paradigm. We then show that the inside behavior of the observer is a source of information as well as the physical properties of visibles surfaces. Examples are proposed in the particular case of connected components extraction in a binary image using a data flow pyramid computer showing that "how" we process information may be more important than "what" we actually process.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 03.95
Keywords: Pyramid, Behavioral model, Temporal analysis
Christophe Duperthuy and Jean-Michel Jolion
Category: Conference
Abstract: Our goal in this paper is to extend usual models of visual data to be extracted from pictures, i.e. generalizing the primal sketch. We suggest here a binary approach based on features' continuity and directionality, using the contrasts pyramid. This yields a set of attributes maps, such as noises, textures, edges and blobs, over eight basic operations.
Publication-Status: 8th Int. Workshop on Theoretical Foundations of Computer Vision, Dagstuhl, Germany, March 1996, Springer Verlag, 1997
Keywords: Primal sketch, binary mask, pyramid, contrast
Jean-Gérard Pailloncy and Jean-Michel Jolion
Category: Report
Abstract: We have the habit to consider a region as a single connexe colored zone of an image. In this paper, we generalize the concept of region. We propose a point of view where a generalized region is a function from the same support than the image to [0,1]. We can consider this function as a probability of pixel's membership or as a percentage of pixel's membership. We show why this approach is interesting, how to compute the characteristics of this layer. An application in the irregular pyramid is proposed.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 05.96
Keywords: Image model, cellular complex, region-based models, irregular pyramid
Christophe Duperthuy and Jean-Michel Jolion
Category: Report
Abstract: Most of time, image processing involves a feature-based representation of the picture, suited to its treatment. In this way, Marr suggested a representation of a scene, based on light discontinuities: this is the primal sketch. We attept her to adopt a more general approach to picture representation, that is to extract a set of visual data, beyond light discontinuities.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 07.96
Keywords: Primal sketch, pyramidal multiresolution, early vision, features and orientations extraction
Christophe Duperthuy and Jean-Michel Jolion
Category: Report
Abstract: Commonly, image analysis is based on features such as edges or regions... In this paper, we explore a quite new approach, accounting directional information in order to realize perceptive tasks. This leads to a model based on 3 x 3 binary masks for binary pictures, performing some suprising discriminations.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 09.96
Keywords: Orientations extraction, early vision, primal sketch, pyramidal multiresolution
Jean-Michel Jolion
Category: Report
Abstract: We present in this paper a study of concavity detection using binary edge mask information, and consistency links between them. A model for a corner is proposed as a particular arrangement of edge masks. In a second step, we merge these corners in order to build concavities. Enclosures are obtained as a particular case of concavities. Examples are shown in the context of suburban area and more particularly in house images. We also discuss about the limits of this approach in terms of information provided by the edge masks and propose some improvements in order to extend our concavity definition.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 01.97
Keywords: Corner, concavity, shape representation, symbolic analysis
Jean-Gérard Pailloncy and Jean-Michel Jolion
Category: Conference
Abstract: In this paper, a topologically consistent representation for images is presented. In addition, this representation is suitable for parallel computers, and may easily extend to a pyramid of representations. It is based on the cell complex theory; An efficient algorithm to build the graph is presented.
Publication-Status: GbR'97, 1st Intern. Workshop on Graph based Representations, April 15-17, Lyon, France.
Keywords: Image model, cellular complex, region-based models, irregular pyramid
Jean- Michel Jolion, Jean-Yves Buffiere and Catherine Verdu
Category: Conference
Abstract: In this paper, we use the classic bayesian model to show that one needs a perceptive model when desinging an artificial image analysis system for quality control purposes in order to take into account the way the observer looks at a scene. An exemple in microheterogeneous material interpretation is given.
Publication-Status: International Conference on Quality Control by Artificial Vision, QCAV'97, Le Creusot, France, May 28-30, 1997, 103-108.
Keywords: perceptive model, bayesian model, quality control, pyramid, material image interpretation
Christophe Duperthuy and Jean-Michel Jolion
Category: Conference
Abstract: Nous abordons dans ce papier le problème de la simplification extrème d'une image numérique couleur, suivant un principe de rehaussement du contraste multirésolution: la représentation simplifiée que nous obtenons (mçeme grossière) peut alors s'avérer utile pour un transfert ou un traitement de l'image, du type de ce qui est nécessaire dans les applications d'indexation et de recherche dans les banques d'images.
Publication-Status: 16ème Colloque Gretsi sur le traitement du signal et des images, Grenoble, France, 15-19 Septembre 1997.
Keywords: Contraste, couleur, simplification d'images
Jean-Michel Jolion and Stéphane Bres
Category: Report
Abstract: This report (in french) propose a study of robustness of commonly used imagfe descriptors (like those based on differentials) against coding noise and more particularly the one related to Jpeg.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 01.98
Keywords: Jpeg, image descriptors, coding noise, image indexation.
Stéphane Bres and Jean-Michel Jolion
Category: Report
Abstract: This paper addresses the problem of detection and delineation of interest points in image and sequence of images. This study is part of our current research in the field of automatic image and video indexing for search by content purposes. In the field of image and video indexing, one is often interested in compact features extracted from the signal. More particularly, one of the most popular approach to large image database search is iconic request, e.g. find some images similar to the one given as example. Some now wellknown products are available but they are not so powerful especially because nobody really know what "similar" means. A recently introduced approach is based on interest points. It argues that two signals are similar if they have particular characteristics spatially located in a consistent order. The locations of these particular characteristics are called the interest point or key points. In this paper, we will first present a differential approach of key points detection and delineation. We thus introduce our model based on the multiresolution contrast energy and compare both approaches on images. Last section of the paper discusses current results and present further studies that have to be carried on in order to better emphasize this novel approach.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 02.98
Keywords: Interest points, multiresolution, contrast, pyramid.
Jean-Michel Jolion
Category: Report
Abstract: Le domaine récent de l'indexation d'images a fait
émerger de nombreux nouveaux problèmes pour la communauté du traitement
d'images te de la vision artificielle. Déjà, certains blocages apparaissent
et tout particulièrement dans le domaine des requètes par l'exemple.
Sont-ils nouveaux ? Nous montrons dans ce papier qu'il y a bien des
ressemblances avec des questions restées sans réponses par le passé.
Nous proposons, à la lumière d'une mise en correspondance "historique"
des pistes pour les recherches à mener dans les années futures en
nous appuyant sur notre expérience acquise au sien du projet SESAME.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 05.98
Keywords: Indexation d'images, Paradigme de Marr, Recherche par l'exemple.
Stéphane Bres and Jean-Michel Jolion
Category: Conference
Abstract: This paper addresses the problem of detection and delineation of interest points in images as part of an automatic image and video indexing for search by content purposes project. We propose a novel key point detector based on multiresolution contrast information. We compare this detector to the Plessey feature point detector as well as the detector introduced in the SUSAN project. As we are interested in common database applications, we focus this comparison on robustness vs coding noise like Jpeg noise.
Publication-Status: 3rd Int. Conf. on Visual Information System, Visual99,
2-4 June 1999, Amsterdam, The Netherlands
Keywords: Interest points, multiresolution, contrast, pyramid.
Alexander Heinrichs, Dimitri Koubaroulis, Barbara Levienaise-Obadia, Paulo Rovida and Jean-Michel Jolion
Category: Report
Abstract:
In this report, we focus on robust image retrieval. We first describe
why robustness is of importance in this kind of search. Then we
propose a very general architecture based on three steps. As an
assumption, we assume that any
image of the database is characterized by a set of features (each
feature resulting in a statistical distribution of a particular
measure over the image). The first step of the process consists in
computing, for any characteristic, a similarity/dissimilarity
measure between the distribution related to the image request and those
of the images in the database. This results in a set of scalar
values that have to be summarize in a 1D value in order to easily
sort the image with respect to the request. This is what the second step
is about. Even when powerful tools (i.e. robust...) are
used, given the m best matches, it can happen that part of
the answers is not satisfactory. Therefore the third step intends
to increase the robustness of the overall process using a feedback
analysis (which can be supervised or non supervised). This step
mainly tries to emphasize on the role of the more discriminant
characteristics.
We will show in this report some approaches to improve the robustness
of these three steps.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 04.99
Keywords: Image Indexation, Robust statistics
Jean-Gérard Pailloncy, Aline Deruyver and Jean-Michel Jolion
Category: Conference
Abstract: The RAG has been widely used to represent the topology of the image
segmentation. The RAG has only one type of edges "to be neighbor
with" and its simplicity allows us to easily build a hierarchy of
RAGs. But it is too simple. At the opposite, the semantic graph has
many types of edges with complex properties and it can embed some
very high informations. But they are build by hand due to this
complexity. We will focus on the middle position between the RAG and
the semantic graph and develop some new ideas.
Publication-Status: GbR'99, 2nd Intern. Workshop on Graph based Representations, May, Hainsdorf, Autriche.
Keywords:Image model, RAG, semantic graph, hierarchical graph
Jean-Michel Jolion
Category: Report
Abstract:
The Benford's law has been proposed in the very past in order to
modelize the probability distribution of the first digit in a set
of natural numbers. In this report, we prove that this law is
validated in the digital images domain, not for the images but for
the magnitude of their gradient. We show, experimentaly, that
this also apply for the laplacian pyramid code. This yields to
the field of entropy based coding which takes advantage of any
a priori information about the probability of any symbol
in the signal.
Publication-Status: Laboratoire Reconnaissance de Formes et Vision, technical report, RR 01.00
Keywords: Image Coding, Gradient pdf
Jean-Michel Jolion
Category: Book
Abstract:
This book presents the basic principles of computer vision system design. The basic techniques relative to image processing are not included. The retained point of view is the system oriented approach. Contain: survey on the human visual system, methodology, image sensing, high performances architecture, software architecture for control and learning, images and decision, evaluation, industrial applications.
Publication-Status: Copyright HERMES Sc. Pub, 2000
Keywords: Computer Vision, Human Vision, Architecture, Parallelism, Applications, Methodology, Sensors.
Size: 380p