Consciously or unconsciously, humans always pay attention to a wide variety of stimuli. Attention is part of daily life and it is the first step to understanding. The proposed thesis deals with a computational approach to the human attentional mechanism and with its possible applications mainly in the field of computer vision. In a first stage, the text introduces a rarity-based three-level attention model handling monodimensional signals as well as images or video sequences. The concept of attention is defined as the transformation of a huge acquired unstructured data set into a smaller structured one while preserving the information: the attentional mechanism turns rough data into intelligence. Afterwards, several applications are described in the fields of machine vision, signal coding and enhancement, medical imaging, event detection and so on. These applications not only show the applicability of the proposed computational attention method, but they also support the idea that similarly to the fact that attention is the beginning of intelligence in humans, computational attention may be the starting point of artificial intelligence in engineering applications. Several databases containing different kinds of signals were used to test the model and its applications: audio signals of natural complex ambiences and events, real-life video sequences as well as simulated sequences and finally natural scenes, textured or synthetic images. Results are presented in a clear and comprehensive way within each application providing the relevance of the use of the computational attention model. Finally, a large discussion is opened based on the theoretical and practical achievements and future extensions are proposed.