Analysis and synthesis (A/S) approaches are common in multimedia signal processing. Reducing a signal to its constituent parts is relevant because once the signal is analyzed, the set of extracted parameters can be more convenient for storage, finite length representation, transmission or manipulation than the signal in its original form. The community of audio artists have adopted A/S techniques as an important part of their arsenal of technological tools for the creative manipulation of sounds. Parameters can be altered before the synthesis, generating sophisticated sound effects while maintaining the identity of the original signal. Tools such as the tracking phase vocoder allow musicians to manipulate pitch and duration independently or to combine two signals in exotic ways (cross-synthesis). However, a similar emphasis has yet to be pursued for the creative manipulation of video signals. This dissertation describes how A/S approaches can be used to obtain non-photorealistic representations from live video. In this research, new algorithms have been designed to overcome problems like temporal coherence and to extend concepts like the cross-synthesis of two video signals. The creative and narrative possibilities of the developed techniques are used to compose complete pieces divided between short animations and real time installations. Details on the design and implementation of those pieces are included in this document. The pieces, together with the different examples used throughout the dissertation illustrate the versatility and control that can be obtained with A/S approaches and confirm that these sets of techniques can be considered for the creative manipulation of video data.