Le codage de source distribué est un paradigme qui consiste à encoder indépendamment deux sources corrélées et à les décoder conjointement. Wyner et Ziv ont montré que le codage de source distribué peut atteindre les mêmes performances débit-distorsion que le codage de source conjoint, pourvu que certaines contraintes soient satisfaites. Cette caractéristique rend le codage de source distribué très attractif pour des applications qui demandent un encodeur à faible complexité ou pour ne pas être obligé à avoir des communications entre les sources. Dans le cadre du codage vidéo distribué, les trames corrélées sont encodées séparément et décodées conjointement. Dans l'architecture ainsi dite de Stanford, le flux vidéo est séparée en trames clés et Wyner-Ziv. Les trames clés sont encodées INTRA. Les trames Wyner-Ziv sont données en entrée à un codeur de canal systématique ; seulement les bits de parité sont envoyés. Au décodeur, on produit une estimation de la trame Wyner-Ziv, appelée information adjacente, en interpolant les trames clés reçues. L'information adjacente, considérée comme une version bruitée de la trame Wyner-Ziv, est corrigée par les bits de parité. Dans cette thèse, nous proposons plusieurs algorithmes pour la génération de l'information adjacente et pour l'interpolation temporelle et inter-vue. On propose aussi un algorithme de fusion bayésienne des deux interpolations. Tous les algorithmes proposés donnent des résultats meilleurs par rapport à l'état de l'art en termes de performance débit-distorsion. Nous proposons aussi plusieurs algorithmes pour l'estimation de la trame Wyner-Ziv dans le cadre de la vidéo multi-vues plus profondeur. / Distributed source coding is a paradigm that consists in encoding two correlated sources independently, provided that they are decoded jointly.Wyner and Ziv proved that distributed source coding can attain the same rate distortion performance of joint coding, under some constraints.This feature makes distributed source coding very attractive for applications that require a low-complexity encoder or for avoiding communication between the sources. In distributed video coding, correlated frames are encoded separately but decoded jointly. In the Stanford Architecture, the video is split into Key Frames and Wyner-Ziv Frames. The Key Frames are INTRA coded. The Wyner-Ziv Frames are fed into a systematic channel coder and only the parity bits are sent to the decoder. At the decoder side, an estimation of the Wyner-Ziv Frame, called side information, is produced by interpolating the available frames. The side information, that can be considered as a noisy version of the real Wyner-Ziv Frame, is corrected by the parity bits sent by the encoder. In this thesis, we propose several algorithms for side information generation both for the temporal and inter-view interpolation. We also propose a Bayesian fusion of the two estimations. All our algorithms outperform the state-of-the-art in terms of rate distortion performance. We also propose several algorithms for Wyner-Ziv estimation in the context of multiview video plus depth.

Current interframe video compression standards such as the MPEG4 and H.264, require a high-complexity encoder for predictive coding to exploit the similarities among successive video frames. This requirement is acceptable for cases where the video sequence to be transmitted is encoded once and decoded many times. However, some emerging applications such as video-based sensor networks, power-aware surveillance and mobile video communication systems require computational complexity to be shifted from encoder to decoder. Distributed Video Coding (DVC) is a new coding paradigm, based on two information-theoretic results, Slepian-Wolf and Wyner-Ziv, which allows exploiting source statistics at the decoder only. This architecture, therefore, enables very simple encoders to be used in video coding. Wyner-Ziv video coding is a particular case of DVC which deals with lossy source coding where side information is available at the decoder only. In this thesis, we implemented a DVC codec based on the DISCOVER (DIStributed COding for Video sERvices) project and carried out a detailed analysis of each block. Several algorithms have been implemented for each block and results are compared in terms of rate-distortion. The implemented architecture is aimed to be used as a testbed for future studies.

Vision systems have become ubiquitous. They are used for traffic monitoring, elderly care, video conferencing, virtual reality, surveillance, smart rooms, home automation, sport games analysis, industrial safety, medical care etc. In most vision systems, the data coming from the visual sensor(s) is processed before transmission in order to save communication bandwidth or achieve higher frame rates. The type of data processing needs to be chosen carefully depending on the targeted application, and taking into account the available memory, computational power, energy resources and bandwidth constraints. In this dissertation, we investigate how a vision system should be built under practical constraints. First, this system should be intelligent, such that the right data is extracted from the video source. Second, when processing video data this intelligent vision system should know its own practical limitations, and should try to achieve the best possible output result that lies within its capabilities. We study and improve a wide range of vision systems for a variety of applications, which go together with different types of constraints. First, we present a modulo-PCM-based coding algorithm for applications that demand very low complexity coding and need to preserve some of the advantageous properties of PCM coding (direct processing, random access, rate scalability). Our modulo-PCM coding scheme combines three well-known, simple, source coding strategies: PCM, binning, and interpolative coding. The encoder first analyzes the signal statistics in a very simple way. Then, based on these signal statistics, the encoder simply discards a number of bits of each image sample. The modulo-PCM decoder recovers the removed bits of each sample by using its received bits and side information which is generated by interpolating previous decoded signals. Our algorithm is especially appropriate for image coding. / Morbee, M. (2011). Optimized information processing in resource-constrained vision systems. From low-complexity coding to smart sensor networks [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/12126 / Palancia