Velotiaray Toto-Zarasoa

• December 2010 - May 2011: Expert Engineer in the TEMICS project. Subject: Scalable and Robust Wavelet-based Video coding.
• October 2007 - November 2010: PhD Thesis in the TEMICS project (IRISA), on the subject of Distributed Video Coding.
• October 2004 - September 2007: Master Degree in Computer Science at the IFSIC, Université de Rennes 1.
• September 2002 - July 2004: 2 year selective, preparatory classes leading to French elite Institutes.

My complete Resume is here.

PhD Thesis

• Title: Distributed Source Coding: Tools and application to Video Compression
  (Codage de Sources Distribué : outils et application à la compression vidéo)
• Advisors: Christine Guillemot and Aline Roumy
• Distributed source coding is a technique that allows to compress several correlated sources, without any cooperation between the encoders, and without rate loss provided that the decoding is joint. Motivated by this principle, distributed video codin has emerged, exploiting the correlation between the consecutive video frames, tremendously simplifying the encoder, and leaving the task of exploiting the correlation to the decoder.
  The first part of our contributions in this thesis presents the asymmetric coding of binary sources that are not uniform. We analyze the coding of non-uniform Bernoulli sources, and that of hidden Markov sources. For both sources, we first show that exploiting the distribution at the decoder clearly increases the decoding capabilities of a given channel code. For the binary symmetric channel modeling the correlation between the sources, we propose a tool to estimate its parameter, thanks to an EM algorithm. We show that this tool allows to obtain fast estimation of the parameter, while having a precision that is close to the Cramer-Rao lower bound.
  In the second part, we develop some tools that facilitate the coding of the previous sources. This is done by the use of syndrome-based Turbo and LDPC codes, and the EM algorithm. This part also presents new tools that we have developed to achieve the bounds of asymmetric and non-asymmetric distributed source coding. We also show that, when it comes to non-uniform sources, the roles of the correlated sources are not symmetric.
  Finally, we show that the proposed source models are well suited for the video bit planes distributions, and we present results that proof the efficiency of the developed tools. The latter tools improve the rate-distortion performance of the video codec in an interesting amount, provided that the correlation channel is additive.

Master Thesis

• Title: Symmetric and rate adaptive coding of correlated sources
  (Codage symétrique et adaptatif en débit de sources corrélées)
• Advisors: Christine Guillemot and Aline Roumy
• The problem of the coding of distributed sources relates to lossless compression of several sources correlated but localised at different places, without possibility of communications between the sources. If the sources are jointly treated the limit of compression is given by first theorem of Shannon: the joint entropy of the sources. On the other hand, if the sources are treated separately (disjoint coding but joint decoding), Slepian and Wolf (1973) show that the compression ratio which can be reached is the same as in the case of a joint coding. In other words, to reach the optimal compression ratio, there is no additional information to transmit if the sources cannot cooperate. Only knowledge of the correlation is necessary to the coding.
  An example of application of Distributed Source Coding (DSC) is a sensors network, where several sensors measure dependent data (temperature for example) and transmit this information in a data fusion center. The problem also arises in the case of video compression, where complexity is deferred to the receiver.
  If the theoretical results are known since nearly 30 years, it is only recently that algorithms of compression were proposed in this context. Then many problems remain open. For example: if the knowledge of the achievements of the other sources is not necessary to code a source, on the other hand the knowledge of the correlation between the sources is necessary to the choice of the algorithm to implement, namely the pair coder/decoder. However this correlation may have been estimated. It is this problem which is tackled in this work. The objective of this training course is thus to study robustness of the existing techniques of DSC to a partial knowledge of the correlation between the sources, to even propose robust algorithms estimating this same law. It is to note that the current solutions for DSC rest on the use of errors correcting codes and the subject of this training course thus contains the study of the robustness of correct codes to a partial knowledge of the characteristics of the channel.

Publications

	V. Toto-Zarasoa and A. Roumy and C. Guillemot Maximum Likelihood BSC parameter estimation for the Slepian-Wolf problem IEEE Communications Letters, vol. 15, no. 2, pp. 232—234, February, 2011.
	V. Toto-Zarasoa and A. Roumy and C. Guillemot Source modeling for Distributed Video Coding IEEE Transactions on Circuits and Systems for Video Technology (accepted for publication), 2011.
	G. Coluccia and E. Magli and A. Roumy and V. Toto-Zarasoa Lossy Compression of Distributed Sparse Sources: a Practical Scheme European Signal Processing Conference (EUSIPCO), 2011.
	V. Toto-Zarasoa and A. Roumy and C. Guillemot Non-uniform source modeling for Distributed Video Coding European Signal Processing Conference (EUSIPCO), pp. 1889—1893, August, 2010.
	V. Toto-Zarasoa and A. Roumy and C. Guillemot Hidden Markov model for Distributed Video Coding IEEE International Conference on Image Processing (ICIP), pp. 3709—3712, September, 2010.
	V. Toto-Zarasoa and A. Roumy and C. Guillemot Non-asymmetric Slepian-Wolf coding of non-uniform Bernoulli sources The 6th International Symposium on Turbo codes (ISTC), pp. 314—318, September, 2010.
	C. Herzet and V. Toto-Zarasoa and A. Roumy Error resilient non-asymmetric Slepian-Wolf Coding IEEE International Conference on Communications (ICC), pp. 1—5, June, 2009.
	V. Toto-Zarasoa and A. Roumy and C. Guillemot and C. Herzet Robust and fast non asymmetric Distributed Source Coding using Turbo Codes on the Syndrome trellis IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 2477—2480, April, 2009.
	V. Toto-Zarasoa and E. Magli and A. Roumy and G. Olmo On Distributed Arithmetic Codes and Syndrome Based Turbo Codes for Slepian-Wolf Coding of Non Uniform Sources European Signal Processing Conference (EUSIPCO), pp. 2249—2253, August, 2009.
	V. Toto-Zarasoa and A. Roumy and C. Guillemot Rate-adaptive codes for the entire Slepian-Wolf region and arbitrarily correlated sources IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pp. 2965—2968, April, 2008.
	V. Toto-Zarasoa Distributed source coding: Tools and application to video compression PhD. thesis, Matisse, November, 2010.