This paper presents (in french) an integrated hardware and software environment dedicated to the prototyping of real-time image processing (IP) applications. It relies on an original programming model that can act simultaneously as a functional expression of algorithms and of their implementation on a target data-flow computer. This computer, built at ETCA, embeds 1024 custom data-flow processors and several transputers in a 3D network. Algorithms, first expressed with a syntax similar to that of Backus'FP language, are converted into data-flow graphs of real-time operators that can be physically mapped onto the processor array. The programming environment includes an interactive data-flow graph compiler, a large library of IP functional operators, an automatic place an route tool along with graphical X-Window based tools for visualizing and debugging network configurations. The whole system allows fast design and prototyping of a wide range of vision applications operating on digital video streams. Among these, we describe here the implementation of a connected component labeling algorithm running on the fly on 25 Hz sequences of 512 x 512 images.