A short history

Development of DIPlib started in 1995, at the capable hands of Geert van Kempen and Michael van Ginkel, under direction of Lucas van Vliet, at the Pattern Recognition Group (later Quantitative Imaging, now Computational imaging) of Delft University of Technology (TU Delft). Most of the algorithms that had been developed there were included in the library, together with a large collection of standard algorithms. Due to the lack of a C++ standard at the time, they developed the library in C, recreating much of the C++ functionality (templates, function overloading, exceptions, data hiding, memory management) using preprocessor macros and other tricks. DIPlib was originally used on HPUX, Solaris and IRIX, and later on Windows, Linux, Mac OS 9 and Mac OS X.

Originally, the DIPlib library was used from within the SCIL_Image image processing software. In 1999, Cris Luengo (with a lot of help from Michael van Ginkel) wrote an interface to MATLAB, defining a flexible and intuitive command-line syntax for the development of image analysis algorithms. That same year, a user-friendly GUI in the spirit of SCIL_Image was written, as well as interactive image display tools. This MATLAB toolbox, called DIPimage, became the primary interface to the DIPlib library.

DIPlib 3 represents the first major rewrite of the DIPlib code base. Planning started in 2014, and by 2017 Cris Luengo had rewritten the infrastructure in C++14, using all the original ideas and concepts, but adding tensor images, color support, and other ideas we had developed within the DIPimage toolbox. C++14 allows the user to write code that is almost as short as the equivalent MATLAB code, making it simple to use the library even for rapid prototyping. The new infrastructure is much easier to read, maintain, and contribute to. The original intention was to make minimal changes to the image analysis routines so that they would work within the new infrastructure, but we ended up significantly rewriting many of those routines, and some were rewritten from scratch to use more efficient algorithms underneath. Bernd Rieger provided financial support for some of this effort through a European Research Council grant. Very few routines have not been ported yet.

The DIPimage toolbox has been updated to optimally use DIPlib 3. This means that some of the MATLAB code was replaced with calls to DIPlib, and the low-level interface (dip_* functions) disappeared. The internal representation of images has also changed. However, we strove to keep backwards-compatibility in the high-level toolbox functions. Some functions have not yet been ported over.

With financial support from Flagship Biosciences, Cris Luengo created the initial Python bindings (PyDIP) in 2017.

Wouter Caarls created the DIPviewer and DIPjavaio components in 2017 and 2019, respectively.

See DIPlib contributors for a list of people that contributed code to the project.