Ipp4r is a Ruby extension that wraps some of the image processing routines from Intel Integrated Performance Primitives library. I wrote it while working at the Intel Research Lab at the faculty of Computational Mathematics and Cybernetics of the Moscow State University.
Ipp4r is written in C and extensively uses preprocessor metaprogramming to automate the generation of boilerplate code. To simplify some of the tasks I wrote a small library for preprocessor metaprogramming similar to Boost.Preprocessor. Note that Ipp4r was last built against Ruby 1.8.6 and it likely won't compile against the latest Ruby sources.
Main features of Ipp4r are:
* Abstraction. Ipp4r offers a high-level Ruby interface to low-level IPP functions.
* Extensibility. Ipp4r is easily extensible, so that new functions can be added without understanding all the trickery of Ruby implementation.
* Performance. Ipp4r is backed by the Intel Integrated Performance Primitives library, so it significantly outperforms all of the available Ruby image processing libraries (at least that was true at the moment when these lines were written).
Example Ruby code:
{{{
require 'ipp4r'
img = Ipp::Image.jaehne(1000, 1000)
img.add_rand_uniform!(0, 0.2).save("random.jpg")
img.filter_median(Ipp::Size.new(15,15)).save("median.jpg")
}}}
In the third line, `jaehne` sample image of size 1000x1000 is created and stored in the `img` variable. In the 4th line a linearly distributed noise is added to `img` and result is saved to `"random.jpg"`. In the 5th line a median filter of size 15x15 is applied to the noisy image and the result is saved to `"median.jpg"`.
It may be not so trivial to build ipp4r on your system. I have included setup-win32.bat file to make your life easier in case you are building ipp4r under win32 using MSVC. I tested it with MSVC 8.0 and it worked OK for me. Please note that you must run setup-win32.bat from Visual Studio Command Prompt, with ruby in your PATH. In my case the command line looked like this:
$ setup-win32.bat
--opencv
-I "C:\Program Files\Intel\IPP\5.3.2.073\ia32\include"
-L "C:\Program Files\Intel\IPP\5.3.2.073\ia32\stublib"
-I "C:\Program Files\OpenCV\otherlibs\highgui"
-I "C:\Program Files\OpenCV\cxcore\include"
-L "C:\Program Files\OpenCV\lib"
setup-win32.bat takes the same command line options as extconf.rb does:
-v, --ipp-version VERSION use VERSION suffix for IPP libraries
-I, --include-dir DIR look for include files in DIR
-L, --lib-dir DIR look for library files in DIR
--[no-]opencv Use OpenCV
-h, --help Show help
OpenCV is used in ipp4r for loading and saving images in different formats. If you aren't using OpenCV, then ipp4r will work correctly with bmp images only. In case you have specified --opencv switch, you must also make sure that you have added corresponding -L and -I options for OpenCV include and library dirs.
If you're working on a non-win32 platform, then you may have to fix extconf.rb - I have tested it on win32 only. I also made sure that GCC 3.4.* and 4.3.* compiles the sources without errors / warnings, so you shouldn't have any problems compiling ipp4r with it.
There is currently no documentation on ipp4r, but the sources are heavily commented in javadoc style, so you can use Doxygen or some other tool to generate documentation. You can also look at some examples in /example subdirectory, which should provide sufficient introduction.
Please direct any questions to the author,
Alexander Fokin <apfokin@gmail.com>