We deal first with the analogy between physical linear systems and convolution equations, and show the theoretical difficulties encountered when solving such equations (existence, uniqueness...). We define precisely the words deconvolution, apodization used by physicists. We then make a survey (not intended to be exhaustive) of deconvolution methods and algorithms currently proposed by severals authors. In the end we study, when possible, the limitations due to various noises appearing at different levels of the measurement and how, in some circumstances, an a priori information may decrease the importance of these limitations.Tout d'abord, nous rappelons les analogies entre systèmes physiques linéaires et équations de convolution. Nous mon...