Optronic image processing

Image processing is a field where mathematics (to lay the foundations for processing), physics (to understand the principles of acquisition), computer science (to automate processing) and the sciences studying human visual perception (as it remains our ultimate reference and objective in terms of analysis) converge. This field has benefited enormously from developments in electronics, calculation capacity and acquisition resources, as well as in data transmission, with very high data rates today. It's commonplace to say that there are images everywhere, generated by sensors of varying degrees of resolution, and for all kinds of uses: telephones, computers, cars, surveillance cameras, satellites, medical operating theatres...

From a researcher's point of view, the term "optronics" associated with image processing is now seen as obsolete, as the methods used in this field are applicable regardless of the context or field considered. Nevertheless, there are still certain professions for which this vocabulary is important: in the military world, and in the world of embedded computing. This article therefore presents an overview of image processing, starting with general notions of image filtering and manipulation, followed by frequency domain approaches and segmentation techniques. Finally, a more substantial development is made on pattern recognition techniques, with keys to understanding point-of-interest detectors and descriptor calculations, now serving as the basis for numerous operations such as image registration, target tracking and pattern recognition.

This article is intended as an introduction to the basics of treatment, and is by no means exhaustive. It is intended to serve as a reference for those wishing to delve deeper into the subject. However, classical methods are often best suited to lightweight processors with reduced computing capacity and high power consumption constraints, and are therefore likely to be found in hardware currently under development.