Abstract : Within a few years only, the amount of video information transmitted across a large range of communication channels has been critically increasing. It is expected, by 2014, that IP traffic will consist, most exclusively, of video data. In mobiles, video traffic is expected to undergo an increase without precedent as well. Despite the ever-increasing throughput of modern transmission channels, these will not be able to sustain such an increase in payload. More than ever, it is essential to improve our ability to compact the video information. Research, for the past 30 years, provided numerous decorrelation tools that reduce the amount of redundancies across both spatial and temporal dimensions in image sequences. To this day, the classical video compression paradigm locally splits the images into blocks of pixels (macroblocks), and processes the temporal axis on a frame by frame basis, without any obvious continuity. Despite very high compression performances (e.g. AVC and forthcoming HEVC standards), one may still advocate the use of alternative approaches. Disruptive solutions have also been proposed, and notably offer the ability to continuously process the temporal axis. However, they often rely on complex tools (e.g. Wavelets, control grids) whose use is rather delicate in practice. This thesis investigates the viability of an alternative representation that embeds features of both classical and disruptive approaches. Its goal is to exhibit the temporal persistence of the textural information, through a time-continuous description. However, it still relies on blocks, mostly responsible for the popularity of the classical approach. Instead of re-initializing the description at each frame, it is proposed to track the evolution of initial blocks taken from a reference image. A block, and its trajectory across time and space, is called a motiontube. An image sequence is then interpreted as a set of motiontubes. Three major problems have been considered within this thesis. At first, motiontubes need to track both continuous and discontinuous displacements and deformations of individual patches of textures. Above all, it is critical for them to evolve as consistently as possible, which will require dedicated regularization mechanisms. Then, a second problem lies in the texture itself and the way it is used to synthesize images: how to handle non-registered and multi-registered areas. Finally, it is essential for a motiontube to be terminated whenever the corresponding patch of texture disappears or cannot be properly tracked any longer, which raises the problem of quality and efficiency assessment. This has a critical influence on the compactness of the representation. Results will eventually show that tubes can effectively be used to represent image sequences, and compare their performances with those of \AVC standard.