When starting to work on an OpenCV project, one should be aware of its historical past. OpenCV has existed for more than 15 years as an open source project, and despite its very dedicated management team that aims to better the library and keep it relevant, some implementations are more outdated than others. Some APIs are left for backward compatibility with previous versions, and others are targeted at specific algorithmic circumstances, all while newer algorithms are added.
Any engineer looking to choose the best performing algorithm for his work should have the tools to inquire about a specific algorithm to see when it was added and what are its origins (for example, a research paper). That is not to suggest that anything new is necessarily better, as some basic and older algorithms are excellent performers, and in most cases there's a clear trade-off between various metrics. For example, a data-driven deep neural network to perform image binarization (turning...