OpenCV is a vast Image processing and computer vision library that offers ease and flexibility for programmers. A good thing about the library is that it's neatly divided into different modules based on functionality ranging from basic image processing to advanced computer vision algorithms. As of current version (2.4.6), there are a total of 21 modules in the package.

While the modular approach is good to keep the dependencies minimal, some programmers also find it frustrating to link each of the required module separately with the project. When shipping an end-user-application built with OpenCV, all the libraries needed have to be shipped and some developers are not satisfied with this clutter of shared libraries. This is where OpenCV world module comes in.

The "world module" is an all-in-one module that has the functionality of all the libraries. If you're not using pre-built libraries, the 'world module' will have the functionality of all the modules selected while building your own OpenCV from source. The module is only for experimental basis with the current release and is not included in the pre-built binaries shipped with the package. To use the world module for now, you have to build OpenCV from the source, with Cmake option BUILD_opencv_world checked.

Once built, developer just have to link opencv_worldXXX.lib (Windows) or libopencv_world (UNIX), and the project can use any and all of the functionality available with OpenCV. This may look like a quick dirty approach, but in fact it's a preferred way if you want to build end-user applications that require binaries to be shipped. Instead of many shared libraries, all you have to do is ship one. The only small down side of using the world module is you may find a slight increase in compilation time of your application.

The "world module" is expected to become official in the next release (version 2.4.7).

I definitely recommend using world module in your applications, especially when you want to redistribute binaries with end-user applications.