Seismic exploration utilizes controlled sources, which emit seismic waves that propagate through the earth subsurface and get reflected off subsurface interfaces and scatterers. The reflected and scattered waves are recorded by recording stations installed along the earth surface or down boreholes. Seismic imaging is a powerful tool to map these reflected and scattered energy back to their subsurface scattering or reflection points. Seismic imaging is conventionally based on the single-scattering assumption, where only energy that bounces once off a subsurface scatterer and recorded by a receiver is projected back to its subsurface position. The internally multiply scattered seismic energy is considered as unwanted noise and is usually suppressed or removed from the recorded data. Conventional seismic imaging techniques yield subsurface images that suffer from low spatial resolution, migration artifacts, and acquisition fingerprint due to the limited acquisition aperture, number of sources and receivers, and bandwidth of the source wavelet. Hydrocarbon traps are becoming more challenging and considerable reserves are trapped in stratigraphic and pinch-out traps, which require highly resolved seismic images to delineate them. This thesis focuses on developing and implementing new advanced cost-effective seismic imaging techniques aiming at enhancing the resolution of the migrated images by exploiting the sparseness of the subsurface reflectivity distribution and utilizing the multiples that are usually neglected when imaging seismic data. I first formulate the seismic imaging problem as a Basis pursuit denoise problem, which I solve using an L1-minimization algorithm to obtain the sparsest migrated image corresponding to the recorded data. Imaging multiples may illuminate subsurface zones, which are not easily illuminated by conventional seismic imaging using primary reflections only. I then develop an L2-norm (i.e. least-squares) inversion technique to image internally multiply scattered seismic waves to obtain highly resolved images delineating vertical faults that are otherwise not easily imaged by primaries. Seismic interferometry is conventionally based on the cross-correlation and convolution of seismic traces to transform seismic data from one acquisition geometry to another. The conventional interferometric transformation yields virtual data that suffers from low temporal resolution, wavelet distortion, and correlation/convolution artifacts. I therefore incorporate a least-squares datuming technique to interferometrically transform vertical-seismic-profile surface-related multiples to surface-seismic-profile primaries. This yields redatumed data with high temporal resolution and less artifacts, which are subsequently imaged to obtain highly resolved subsurface images. Tests on synthetic examples demonstrate the efficiency of the proposed techniques, yielding highly resolved migrated sections compared with images obtained by imaging conventionally redatumed data. I further advance the recently developed cost-effective Generalized Interferometric Multiple Imaging procedure, which aims to not only image first but also higher-order multiples as well. I formulate this procedure as a linearized inversion framework and solve it as a least-squares problem. Tests of the least-squares Generalized Interferometric Multiple imaging framework on synthetic datasets and demonstrate that it could provide highly resolved migrated images and delineate vertical fault planes compared with the standard procedure. The results support the assertion that this linearized inversion framework can illuminate subsurface zones that are mainly illuminated by internally scattered energy.

La séparation de signaux corrélés à partir de leurs combinaisons linéaires est une tâche difficile et possède plusieurs applications en traitement du signal. Nous étudions deux problèmes, à savoir la séparation aveugle de sources parcimonieuses et le filtrage adaptatif des réflexions multiples en acquisition sismique. Un intérêt particulier est porté sur les mélanges convolutifs : pour ces deux problèmes, des filtres à réponses impulsionnelles finies peuvent être estimés afin de récupérer les signaux désirés.Pour les modèles de mélange instantanés et convolutifs, nous donnons les conditions nécessaires et suffisantes pour l'extraction et la séparation exactes de sources parcimonieuses en utilisant la pseudo-norme L0 comme une fonction de contraste. Des équivalences entre l'analyse en composantes parcimonieuses et l'analyse en composantes disjointes sont examinées.Pour la soustraction adaptative des réflexions sismiques, nous discutons les limites des méthodes basées sur l'analyse en composantes indépendantes et nous soulignons l'équivalence avec les méthodes basées sur les normes Lp. Nous examinons de quelle manière les paramètres de régularisation peuvent être plus décisifs pour l'estimation des primaires. Enfin, nous proposons une amélioration de la robustesse de la soustraction adaptative en estimant les filtres adaptatifs directement dans le domaine des curvelets. Les coûts en calcul et en mémoire peuvent être atténués par l'utilisation de la transformée en curvelet discrète et uniforme. / The recovery of correlated signals from their linear combinations is a challenging task and has many applications in signal processing. We focus on two problems that are the blind separation of sparse sources and the adaptive subtraction of multiple events in seismic processing. A special focus is put on convolutive mixtures: for both problems, finite impulse response filters can indeed be estimated for the recovery of the desired signals.For instantaneous and convolutive mixing models, we address the necessary and sufficient conditions for the exact extraction and separation of sparse sources by using the L0 pseudo-norm as a contrast function. Equivalences between sparse component analysis and disjoint component analysis are investigated.For adaptive multiple subtraction, we discuss the limits of methods based on independent component analysis and we highlight equivalence with Lp-norm-based methods. We investigate how other regularization parameters may have more influence on the estimation of the desired primaries. Finally, we propose to improve the robustness of adaptive subtraction by estimating the extracting convolutive filters directly in the curvelet domain. Computation and memory costs are limited by using the uniform discrete curvelet transform.