This thesis is dedicated to the design of a single-frame passive imaging system able to produce a depth map of the observed scene. This design is based on the joint optimisation of the optical and processing parameters, also it is referred to as codesign. The depth estimation ability of a single-frame passive imaging system relies on the depth from defocus principle (DFD) where depth is estimated based on the local estimation of the defocus blur. A new local depth estimation algorithm is developed. It is unsupervised and generic enough to be used with a large family of imaging systems. Then two optical concepts improving depth estimation accuracy are theoretically and experimentally investigated: a coded aperture or a lens with longitudinal chromatic aberration. The realisation of a chromatic imaging system with depth estimation ability is new and this work underlines the advantages of the chromatic solution, in terms of accuracy and range where depth can be estimated. Then a performance model is developed in order to predict the theoretical depth estimation accuracy of any imaging system that uses DFD. This model predicts the accuracy of an imaging system based on the knowledge of the optical and the processing parameters. It is then used as a tool for the design of the first codesigned chromatic imaging system optimised for depth estimation. The realisation of this prototype has highlighted the interest of the codesign approach.
