cv2 vs. "opencv3"
To get a potential misunderstanding out of the way:
The python OpenCV module is named and imported via import cv2 in all versions > 2.0, including > 3.0. If you want to work with cv2, installing OpenCV versions > 3 is fine - unless you're looking for specific compatibility with older versions or are a fan of the 2.4.x versions. The switch from 2.4.x to 3.x was in 2015 and in terms of features, speed and transparency, it makes much sense to use the newer versions. You can read here and here about major differences. 2.4.x versions are still supported though, current release is 2.4.13.5.
Installing a specific version, e.g. OpenCV 2.4.9
That said:
If you want to install a specific version that neither pip install opencv-python==2.4.X, sudo apt-get install opencv nor conda install opencv=2.4.x provide (as explained by other answers here), you can always install from sources. In the sourceforge repository you can find all major versions for each operating system. Although for unxeperienced users this might be scary, it is well explained in some tutorials. E.g. here for 2.4.9 on Ubuntu 14.04. Or here is the official Linux install doc for the latest release 2.4.13.5.
In essence, the install process boils down to:
install dependencies, refer to docs (e.g. here) for required packages
get sources from OpenCVs sourceforge
e.g. wget http://sourceforge.net/projects/opencvlibrary/files/opencv-unix/2.4.9/opencv-2.4.9.zip
unzip sources and prepare build by creating build directory and running cmake
mkdir build
cd build
cmake (... your build options ...)
build in the created build directory with:
make
sudo make install