Dans cette thèse, nous nous intéressons aux méthodes de la fusion d'images multi focales. Cette technique consiste à fusionner plusieurs images capturées avec différentes distances focales de la même scène. Cela permet d'obtenir une image de meilleure qualité à partir des deux images sources. Nous proposons une méthode de fusion d'images s'appuyant sur les techniques des pyramides Laplaciennes en utilisant comme règle de sélection les transformées d'ondelettes discretes(DWT: Discrete Wavelet Transform). Nous développons, par la suite, deux méthodes de fusion d'images multi focales basée sur la variabilité locale de chaque pixel. Elle tient en compte les informations dans la région environnante des pixels. La première consiste à utiliser la variabilité locale comme information dans la méthode de Dempster-Shafer. La seconde utilise une métrique basée sur la variabilité locale. En effet, la fusion proposée effectue une pondération de chaque pixel par une exponentielle de sa variabilité locale. Une étude comparative entre les méthodes proposées et celles existantes a été réalisée. Les résultats expérimentaux démontrent que nos méthodes proposées donnent des meilleurs fusions, tant dans la perception visuelle que dans l'analyse quantitative. / In this thesis, we are interested in the multi-focus image fusion method. This technique consists of fusing several captured images with different focal lengths of the same scene to obtain an image with better quality than the two source images. We propose an image fusion method based on Laplacian pyramid technique using Discrete Wavelet Transform (DWT) as a selection rule. We then develop two multi-focus image fusion methods based on the local variability of each pixel. It takes into account the information in the surrounding pixel area. The first method is to use local variability as an information in the Dempster-Shafer theory. The second method uses a metric based on local variability. Indeed, the proposed fusion method weighs each pixel by an exponential of its local variability. A comparative study between the proposed methods and the existing methods was carried out. The experimental results show that our proposed methods give better fusions, both in visual perception and in quantitative analysis.

This thesis presents the novel use of Coherent Point Drift in improving the robustness of a number of computer vision applications. CPD approach includes two methods for registering two images - rigid and non-rigid point set approaches which are based on the transformation model used. The key characteristic of a rigid transformation is that the distance between points is preserved, which means it can be used in the presence of translation, rotation, and scaling. Non-rigid transformations - or affine transforms - provide the opportunity of registering under non-uniform scaling and skew. The idea is to move one point set coherently to align with the second point set. The CPD method finds both the non-rigid transformation and the correspondence distance between two point sets at the same time without having to use a-priori declaration of the transformation model used. The first part of this thesis is focused on speaker identification in video conferencing. A real-time, audio-coupled video based approach is presented, which focuses more on the video analysis side, rather than the audio analysis that is known to be prone to errors. CPD is effectively utilised for lip movement detection and a temporal face detection approach is used to minimise false positives if face detection algorithm fails to perform. The second part of the thesis is focused on multi-exposure and multi-focus image fusion with compensation for camera shake. Scale Invariant Feature Transforms (SIFT) are first used to detect keypoints in images being fused. Subsequently this point set is reduced to remove outliers, using RANSAC (RANdom Sample Consensus) and finally the point sets are registered using CPD with non-rigid transformations. The registered images are then fused with a Contourlet based image fusion algorithm that makes use of a novel alpha blending and filtering technique to minimise artefacts. The thesis evaluates the performance of the algorithm in comparison to a number of state-of-the-art approaches, including the key commercial products available in the market at present, showing significantly improved subjective quality in the fused images. The final part of the thesis presents a novel approach to Vehicle Make & Model Recognition in CCTV video footage. CPD is used to effectively remove skew of vehicles detected as CCTV cameras are not specifically configured for the VMMR task and may capture vehicles at different approaching angles. A LESH (Local Energy Shape Histogram) feature based approach is used for vehicle make and model recognition with the novelty that temporal processing is used to improve reliability. A number of further algorithms are used to maximise the reliability of the final outcome. Experimental results are provided to prove that the proposed system demonstrates an accuracy in excess of 95% when tested on real CCTV footage with no prior camera calibration.