Random Thoughts

from and to 372433 143758 48 N

Qui suis-je ?

English

Biographie

Emmanuel Christophe est né en France en 1980. Il a recu son diplôme d’ingénieur en Télécommunication (ENST Bretagne) et son DEA en traitement du signal et des images (Universitée de Rennes 1) avec mention en 2003. Il a travaillé six mois pour la National University of Singapore au sein du laboratoire CWAIP sur la compression video. En octobre 2006, il a recu le diplone de docteur de Supaero grace à son travail sur la compression et la qualité des images hyperspectrales réalisé au TéSA en coopération avec le Centre National d’Etudes Spatiales (CNES), l’Office National d’Etudes et de Recherches Aérospatiales (ONERA) et Alcatel Alenia Space. Il a été en visite au Rensselaer Polytechnic Institute, Troy, NY, USA en 2006. Il est maintenant ingénieur de recherche au CNES. Ses intérêts comprennent la compression d’images et de video, l’évaluation de qualité et les technologies de télédétection.

Activités de recherche

  • Thèse sur la compression des images hyperspectrales.
  • Participation au projet ACI2M : traitement de masse de données, projet impliquant plusieurs laboratoires.
  • Membre du GdR-ISIS: un groupe de recherche sur information, signal, images et vision.
  • Membre de IEEE Society.
  • Volontaire pour l’organisation de ICASSP 2006, Toulouse, mai 2006.
  • Participation active à l’Orfeo Toolbox (OTB, OTB Live)

Enseignements

  • Insa Toulouse: Compression des images satellites (étudiants de 4ème année) (cours).
  • Enseeiht: Introduction au traitement du signal et analyse spectrale.
  • Ensica: Filtrage numérique.
  • Cnam: Cours signal et bruit.

Centres d’intérêt

Divers

  • Distinction IEEE Geoscience and Remote Sensing Society pour le Data Fusion Contest 2008

Publications

Articles de journaux

  1. E. Christophe, C. Mailhes, and P. Duhamel, "Hyperspectral Image Compression: Adapting SPIHT and EZW to Anisotropic 3D Wavelet Coding," IEEE Transactions on Image Processing, vol. 17, iss. 12, pp. 2334-2346, 2008.
    Hyperspectral images present some specific characteristics that should be used by an efficient compression system. In compression, wavelets have shown a good adaptability to a wide range of data, while being of reasonable complexity. Some wavelet-based compression algorithms have been successfully used for some hyperspectral space missions. This paper focuses on the optimization of a full wavelet compression system for hyperspectral images. Each step of the compression algorithm is studied and optimized. First, an algorithm to find the optimal 3D wavelet decomposition in a rate-distortion sense is defined. Then, it is shown that a specific fixed decomposition has almost the same performance, while being more useful in terms of complexity issues. It is shown that this decomposition significantly improves the classical isotropic decomposition. One of the most useful properties of this fixed decomposition is that it allows the use of zero tree algorithms. Various tree structures, creating a relationship between coefficients, are compared. Two efficient compression methods based on zerotree coding (EZW and SPIHT) are adapted on this near-optimal decomposition with the best tree structure found. Performances are compared with the adaptation of JPEG 2000 for hyperspectral images on six different areas presenting different statistical properties.
  2. E. Christophe and W. A. Pearlman, "Three-dimensional SPIHT Coding of Volume Images with Random Access and Resolution Scalability," EURASIP Journal on Image and Video Processing, 2008.
    End users of large volume image datasets are often interested only in certain features that can be identified as quickly as possible. For hyperspectral data, these features could reside only in certain ranges of spectral bands and certain spatial areas of the target. The same holds true for volume medical images for a certain volume region of the subject’s anatomy. High spatial resolution may be the ultimate requirement, but in many cases a lower resolution would suffice, especially when rapid acquisition and browsing are essential. This paper presents a major extension of the 3D-SPIHT (set partitioning in hierarchical trees) image compression algorithm that enables random access decoding of any specified region of the image volume at a given spatial resolution and given bit rate from a single codestream. Final spatial and spectral (or axial) resolutions are chosen independently. Because the image wavelet transform is encoded in tree blocks and the bit rates of these tree blocks are minimized through a rate-distortion optimization procedure, the various resolutions and qualities of the images can be extracted while reading a minimum amount of bits from the coded data. The attributes and efficiency of this 3D-SPIHT extension are demonstrated for several medical and hyperspectral images in comparison to the JPEG2000 Multicomponent algorithm.
  3. E. Christophe, P. Duhamel, and C. Mailhes, "Adaptation of zerotrees using signed binary digit representations for 3 dimensional image coding," EURASIP Journal on Image and Video Processing, 2007.
    Zerotrees of wavelet coefficients have shown a good adaptability for the compression of three dimensional images. EZW, the original algorithm using zerotree, shows good performance and was successfully adapted to 3D image compression. This paper focuses on the adaptation of EZW for the compression of hyperspectral images. The subordinate pass is suppressed to remove the necessity to keep the significant pixels in memory. To compensate the loss due to this removal, signed binary digit representations are used to increase the efficiency of zerotrees. Contextual arithmetic coding with very limited contexts is also used. Finally, we show that this simplified version of 3D-EZW performs almost as well as the original one.
  4. E. Christophe, D. Léger, and C. Mailhes, "Quality Criteria Benchmark for Hyperspectral Imagery," IEEE Transactions on Geoscience and Remote Sensing, vol. 43, iss. 09, pp. 2103-2114, 2005.
    Hyperspectral data appear to be of a growing interest over the past few years. However, applications for hyperspectral data are still in their infancy as handling the significant size of the data presents a challenge for the user community. Efficient compression techniques are required, and lossy compression, specifically, will have a role to play, provided its impact on remote sensing applications remains insignificant. To assess the data quality, suitable distortion measures relevant to end-user applications are required. Quality criteria are also of a major interest for the conception and development of new sensors to define their requirements and specifications. This paper proposes a method to evaluate quality criteria in the context of hyperspectral images. The purpose is to provide quality criteria relevant to the impact of degradations on several classification applications. Different quality criteria are considered. Some are traditionnally used in image and video coding and are adapted here to hyperspectral images. Others are specific to hyperspectral data.We also propose the adaptation of two advanced criteria in the presence of different simulated degradations on AVIRIS hyperspectral images. Finally, five criteria are selected to give an accurate representation of the nature and the level of the degradation affecting hyperspectral data.

Conférences internationales

  1. X. Delaunay, E. Christophe, C. Thiebaut, and V. Charvillat, "Best post-transforms selection in a rate-distortion sense," in IEEE International Conference on Image Processing, ICIP’08, 2008.
    This paper deals with the optimization of a new technique of image compression. After the wavelet transform of an image, blocks of coefficients are further linearly decomposed using a basis selected in a dictionary. This dictionary is known by both the encoder and the decoder. This approach is a generalization of the bandelet transform. This paper investigates the problem of the best basis selection. On each block of wavelet coefficients, this selection is made by minimization of a Lagrangian rate-distortion criterion. Theoretical expressions of the optimal Lagrangian multiplier can be computed based on asymptotic hypotheses. A nearly exhaustive search of the optimal Lagrangian multiplier is done for the compression of high resolution satellite images. This numerical study validates the asymptotic theoretical expressions but as well provides a refined expression of the Lagrangian multiplier. At last, the compression results obtained using those different expressions are compared to the optimal compression results obtained with the exhaustive search.
  2. E. Christophe, C. Thiebaut, and C. Latry, "Compression specification for efficient use of high resolution satellite data," in The XXI Congress, The International Society for Photogrammetry and Remote Sensing, ISPRS’08, 2008, pp. 1283-1286.
    For an efficient usage and distribution of high resolution satellite images, several problems need to be solved. The issue comes with the increasing size of these data. Constrains are different than those of the on-board compression, thus different solutions can be selected. For on-board compression, the main constrains are the computational complexity and the rate attainable with qualified space equipments. For on-the-ground compression, computational constraints are not so strong, but particular care is needed to make sure that the chosen format is widely spread and that users will be able to exploit these data easily.
  3. E. Christophe, J. Inglada, and A. Giros, "ORFEO Toolbox: a complete solution for mapping from high resolution satellite images," in The XXI Congress, The International Society for Photogrammetry and Remote Sensing, ISPRS’08, 2008, pp. 1263-1268.
    One of the main objectives of the Orfeo Toolbox (OTB) is the definition and the development of tools for the operational exploitation of the future sub-metric optic and radar images (rapid mapping, tridimensional aspects, change detection, texture analysis, pattern matching, optic and radar complementarities). The purpose of the OTB is to capitalize a methodological know-how in order to adopt an incremental development approach aiming to efficiently exploit the results obtained by research studies. OTB is interesting for all people working in the remote sensing imagery community. Releasing it under an open source licence, CNES hopes to benefit from contributions of many specialists to help grow the practical use of satellite imagery.
  4. E. Christophe, D. Léger, and C. Mailhes, "New Quality Representation for Hyperspectral Images," in The XXI Congress, The International Society for Photogrammetry and Remote Sensing, ISPRS’08, 2008, pp. 315-320.
    Assessing the quality of a hyperspectral image is a difficult task. However, this assessment is required at different levels of the instrument design: evaluation of the signal to noise ratio necessary for a particular application, determining the acceptable level of losses from compression algorithms for example. It has been shown previously that a combination of five quality criteria can provide a good evaluation of the impact of some degradation on applications, such as classification algorithms for example. This paper refines this concept, providing a representation of the degradation which allows predicting the impact on application
  5. E. Christophe and J. Inglada, "Robust Road Extraction for High Resolution Satellite Images," in IEEE International Conference on Image Processing, ICIP’07, 2007.
    Automatic road extraction is a critical feature for an efficient use of remote sensing imagery in most contexts. This paper proposes a robust geometric method to provide a first step extraction level. These results can be used as an initialization for other algorithms or as a starting point for manual road extraction. Results of the extraction are vectorized for GIS integration and for a better interaction with human experts that can refine the results. The algorithm is fast, has very few parameters and is only slightly affected by the image properties (resolution, noise). The algorithm is available in the open-source Orfeo Toolbox.
  6. C. Thiebaut, E. Christophe, D. Lebedeff, and C. Latry, "CNES Studies of On-Board Compression for Multispectral and Hyperspectral Images," in SPIE, Satellite Data Compression, Communications, and Archiving III, 2007.
    Future high resolution instruments planned by CNES for space remote sensing missions will lead to higher bit rates because of the increase in resolution, dynamic range and number of spectral channels for multispectral (up to 16 bands) and hyperspectral (hundreds of bands) imagery. Lossy data compression is then needed, with compression ratio goals always higher and with low-complexity algorithm. For optimum compression performance of such data, algorithms must exploit both spectral and spatial correlation. In the case of multispectral images, CNES (in cooperation with Thales Alenia Space, hereafter TAS) studies have led to an algorithm using a fixed transform to decorrelate the spectral bands, the CCSDS codec compresses each decorrelated band using a suitable multispectral rate allocation procedure. This lowcomplexity decorrelator is adapted to hardware implementation on-board satellite and is under development. In the case of hyperspectral images, CNES (in cooperation with TAS/TeSA/ONERA) studies have led to a full wavelet compression system followed by zerotree coding methods adapted to this decomposition. We are investigating other preprocessors such as Independent Component Analysis which could be used in both approaches. CNES also participates to the new CCSDS Multispectral and Hyperspectral Data Compression Working Group.
  7. E. Christophe, P. Duhamel, and C. Mailhes, "Signed Binary Digit Representation to Simplify 3D-EZW," in IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP’07, 2007.
    Zerotree based coders have shown a good ability to be successfully adapted to 3D image coding. This paper focuses on the adaptation of EZW for the compression of hyperspectral images with reduced complexity. The subordinate pass is removed so that the location of significant coefficients does not need to be kept in memory. To compensate the quality loss due to this removal, a signed binary digit representation is used to increase the efficiency of zerotrees. Contextual arithmetic coding with very limited contexts is also used. Finally, we show that this simplified version of 3D-EZWperforms almost as well as the original one.
  8. E. Christophe and W. A. Pearlman, "Three-dimensional SPIHT Coding of Hyperspectral Images with Random Access and Resolution Scalability," in Fortieth Annual Asilomar Conference on Signals, Systems, and Computers, 2006, pp. 1897-1901.
    With the increase of remote sensing images, fast access to some features of the image is becoming critical. This access could be some part of the spectrum, some area of the image, high spatial resolution. An adaptation of 3D-SPIHT image compression algorithm is presented to allow random access to some part of the image, whether spatial or spectral. Resolution scalability is also available, enabling the decoding of different resolution images from the compressed bitstream of the hyperspectral data. Final spatial and spectral resolutions are chosen independently. From the same compressed bitstream, various resolutions and quality images can be extracted while reading a minimum amount of bits from the coded data. All this is done while reducing the memory necessary during the compression.
  9. E. Christophe, C. Mailhes, and P. Duhamel, "Best Anisotropic 3-D wavelet decomposition in a rate-distortion sense," in IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP’06, 2006, p. ii-17.
    Hyperspectral sensors have been of a growing interest over the past few decades for Earth observation as well as deep space exploration. However, the amount of data provided by such sensors requires an efficient compression system which is yet to be defined. It is hoped that the particular statistical properties of such images can be used to obtain very efficient compression algorithms. This paper proposes a method to find the most suitable wavelet decomposition for hyperspectral images and introduces the possibility of non isotropic decomposition. The decomposition is made by choosing the decomposition that provides an optimal rate-distortion trade-off. The obtained decomposition exhibits better performances in terms of ratedistortion curves compared to isotropic decomposition for high bitrates as well as for low bitrates.
  10. E. Christophe, D. Léger, and C. Mailhes, "Comparison and evaluation of quality criteria for hyperspectral imagery," in SPIE, Image Quality and System Performance II, 2005, pp. 204-213.
    Hyperspectral data appears to be of a growing interest over the past few years. However, applications for hyperspectral data are still in their infancy. Handling the significant size of hyperspectral data presents a challenge for the user community. To enable efficient data compression without losing the potentiality of hyperspectral data, the notion of data quality is crucial for the development of applications. To assess the data quality, quality criteria relevent to end-user applications are required. This paper proposes a method to evaluate quality criteria. The purpose is to provide quality criteria corresponding well to the impact of degradation on end-user applications. Several quality criteria adapted to hyperspectral context are evaluated. Finally, five criteria are selected to give a good representation of the degradation nature and level affecting hyperspectral data.

Thèses

  1. E. Christophe, "Compression des images hyperspectrales et son impact sur la qualité des données," PhD Thesis , 2006.
    Les images hyperspectrales présentent des caractéristiques spécifiques qui demandent à être exploitées par un algorithme de compression efficace. Cette thèse se consacre à la définition d’un système complet de compression pour les images hyperspectrales. En compression, les ondelettes ont montré une bonne efficacité sur des données diverses tout en conservant une complexité raisonnable. De plus, le codage des coefficients d’ondelettes est souvent assuré par des algorithmes basés sur le principe des arbres de zéros qui sont, à l’heure actuelle, parmi les plus efficaces. C’est pourquoi ce travail de thèse s’est intéressé à définir dans un premier temps une décomposition en ondelettes quasi-optimale au sens d’un critère débit-distorsion pour les images hyperspectrales. Dans un deuxième temps, nous proposons une adaptation des méthodes de codage par arbres (EZW, SPIHT), associées à la décomposition obtenue. Les performances sont comparées à une adaptation de JPEG 2000 pour les images hyperspectrales, démontrant l’intérêt des méthodes proposées. D’autre part, les utilisateurs des images hyperspectrales sont souvent intéressés uniquement par certaines caractéristiques de l’image (résolution ou zone) en fonction de l’application. Une adaptation de l’algorithme précédent est réalisée pour permettre l’accès aléatoire afin de décoder uniquement une partie de l’image (en spatial ou en spectral). La progression en résolution est également disponible permettant de décoder des images à différentes résolutions à partir du même train binaire tout en lisant un minimum de bits. Il est également possible de spécifier de manière indépendante les résolutions spatiales et spectrales souhaitées. Enfin, on ne peut parler de compression (avec pertes) sans définir au préalable un critère de distorsion adapté. Nous définissons ainsi un groupe de cinq critères de qualité présentant une bonne complémentarité. Ces cinq critères ont été choisis afin de pouvoir s’assurer que l’algorithme de compression n’entraine pas des dégradations compromettant l’intérêt des données. Une nouvelle méthode d’utilisation de ces cinq critères de qualité montre une bonne aptitude à distinguer différentes dégradations produites sur l’image. Cette méthode, appliquée au nouvel algorithme de compression montre que les dégradations restent faibles pour des débits autour de 1 bit par pixel par bande. Hyperspectral images present some specific characteristics that should be used by an efficient compression system. This thesis focuses on the definition and the optimization of a full wavelet compression system for hyperspectral images. In compression, wavelets have shown a good adaptability to a wide range of data, while being of reasonable complexity. Zerotree based compression algorithms are among the best for image compression. Therefore, in this work, efficient compression methods based on zerotree coding (EZW, SPIHT) are adapted on a near-optimal wavelet decomposition for hyperspectral images. Performances are compared with the adaptation of JPEG 2000 for hyperspectral images. End users of hyperspectral images are often interested only in some specific features of the image (resolution, location) which depend on the application. A further adaptation of the proposed hyperspectral image compression algorithm is presented to allow random access to some part of the image, whether spatial or spectral. Resolution scalability is also available, enabling the decoding of different resolution images from the compressed bitstream of the hyperspectral data while reading a minimum amount of bits from the coded data. Final spatial and spectral resolutions are chosen independantly. Finally, any lossless compression method cannot be characterized without the definition of a distortion measure. Therefore, a group of five quality criteria presenting a good complementarity is defined. The purpose is to make sure the compression algorithm does not impact significantly the data quality. A new method using these five criteria shows a good ability to discriminate between different degradations. Application of this method to the newly defined algorithm shows that the degradation remains low for compression rate around 1.0 bit per pixel per band.

Conférences nationales et workshop

  1. E. Christophe, C. Thiebaut, W. A. Pearlman, C. Latry, and D. Lebedeff, "Zerotree-Based Compression Algorithm for Spaceborne Hyperspectral Sensor," in ESA, On-Board Payload Data Compression Workshop, OBPDC 2008, 2008.
  2. J. -M. Gaucel, E. Christophe, C. Pierangelo, C. Thiebaut, Y. Bobichon, E. Pequignot, F. Lemasson, and D. Lebedeff, "Analysis of Lossy and Lossless Compression Approaches for Futur Ultraspectral Sounder Missions," in ESA, On-Board Payload Data Compression Workshop, OBPDC 2008, 2008.
  3. X. Delaunay, C. Thiebaut, E. Christophe, R. Ruiloba, M. Chabert, V. Charvillat, and G. Morin, "Lossy Compression by Post-Transforms in the Wavelet Domain," in ESA, On-Board Payload Data Compression Workshop, OBPDC 2008, 2008.
  4. J. Inglada, E. Christophe, and H. de Boissezon, "From satellite images to operational applications: generic tools for specific user needs," in Space Appli 2008, 2008.
  5. J. Inglada, E. Christophe, S. Marzocchi-Polizzi, H. de Boissezon, and B. Boissin, "ORFEO Tool Box: an open source library of image processing algorithms for SAR and optical high resolution images," in ESA-EUSC 2008: Image Information Mining: pursuing automation of geospatial intelligence for environment and security, 2008.
  6. E. Christophe, "Fast and robust algorithm for road extraction in high resolution optical images," in 7th CNES/DLR Workshop on Information Extraction and Scene Understanding for Meter Resolution Images, 2007.
  7. E. Christophe, C. Mailhes, and P. Duhamel, "Décorrelation des images hyperspectrales avec une décomposition 3D en ondelettes," in Workshop on Transform Based on Independent Component Analysis for Audio, Video and Hyperspectral Images Data Reduction and Coding, 2006.
    La quantité de données produite par les capteurs hyperspectraux nécessite un algorithme de compression efficace qui reste à définir. Les propriétés statistiques particulières devraient permettre d’obtenir des algorithmes de compression efficaces. Étant données ses propriétés et sa faible complexité, la transformée en ondelettes est un candidat prometteur pour la décorrélation des images hyperspectrales. Ce papier propose une méthode pour trouver la décomposition en ondelettes optimale pour les images hyperspectrales et introduit la possibilité d’une décomposition non isotropique. La décomposition donnant le meilleur compromis débit-distortion est choisie. Cette décomposition donne de bien meilleures performances en terme de débit-distortion que la décomposition isotropique classique. L’inconvénient de cette décomposition optimale réside dans sa complexité importante. Une seconde décomposition, fixe cette fois, est définie et montre des performances quasi optimales tout en gardant une complexité faible.
  8. E. Christophe, D. Léger, and C. Mailhes, "Images hyperspectrales et critères qualité," in Colloque annuel des Doctorants EDIT’05, 2005, pp. 132-136.
    L’intérêt pour les données hyperspectrales est croissant au cours des dernières années. Traiter ces quantités de données importantes présente un défi et la plupart des applications sont encore en développement. Pour définir un système de compression de données efficace sans perdre le potentiel important de ces images, la définition de critères qualité est une étape indispensable. Cet article présente une méthode pour valider des critères qualité en fonction de leur capacité à prédire l’influence de dégradation sur les applications des utilisateurs finaux. Un nombre important de critère qualité est défini, puis évalué. Finalement, cinq critères sont retenus pour donner une bonne estimation de l’impact des dégradations sur des applications classiques de classification.

Les papiers IEEE ci-dessous sont soumis au copyright suivant:
“©20xx IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.”

Les papiers Eurasip sont sous Creative Common Attribution Licence