Servo - January 2007

Math routines for image processing, computer vision, and image interpretation

OpenCV includes math commonly used, algorithms from linear algebra, statistics, and computational geometry.

FIGURE 2. Selecting OpenCV header files in Windows to place into a single include directory.

Graphics

These interfaces let you write text and draw on images. In addition to various fun and creative possibilities, these functions are useful for labeling and marking. For example, if you write a program that detects objects, it’s helpful to label images with their sizes and locations.

experimental code. However, the simplest interfaces for face recognition are in this module. The code behind them is specialized for face recognition, and they’re widely used for that purpose.

GUI methods

OpenCV includes its own windowing interfaces. While these are limited compared to what can be done on each platform, they provide a simple, multi-platform API to display images, accept user input via mouse or keyboard, and implement slider controls.

Datastructures and algorithms

With these interfaces, you can efficiently store, search, save, and manipulate large lists, collections (also called sets), graphs, and trees.

Data persistence

These methods provide convenient interfaces for storing various types of data to disk and retrieving them later.

Figure 1 shows a few examples of OpenCV’s capabilities in action: face detection, contour detection, and edge detection.

Organization

OpenCV’s functionality is contained in several modules.

CXCORE contains basic datatype definitions. For example, the data structures for image, point, and rectangle are defined in cxtypes.h. CXCORE also contains linear algebra and statistics methods, the persistence functions, and error handlers. Somewhat oddly, the graphics functions for drawing on images are located here, as well.

CV contains image processing and camera calibration methods. The computational geometry functions are also located here.

CVAUX is described in OpenCV’s documentation as containing obsolete and

ML contains machine-learning interfaces

The remaining functionality is contained in HighGUI and CVCAM. Both of these are located in a directory named “otherlibs,” making them easy to miss. Since HighGUI contains the basic I/O interfaces, you’ll want to be sure you don’t overlook it! It also contains the multi-platform windowing capabilities.

CVCAM contains interfaces for video access through DirectX on 32-bit Windows platforms. However, HighGUI also contains video interfaces. In this article, I’ll cover only the interfaces in HighGUI. They’re simpler to use, and they work on all platforms. If you’re using Windows XP or 2000, you may get a performance boost by switching to the CVCAM interfaces, but for learning OpenCV, the simpler ones in HighGUI are just fine.

Installing OpenCV

Basic Install

OpenCV for Linux or MacOSX is packaged as a source-code archive. You’ll need to build both static and shared-object libraries. You can either build an RPM first and install from that, or compile and install it directly. Instructions for doing both are in INSTALL.

The Windows download is packaged as an executable that installs OpenCV when you run it. It places OpenCV files into a directory of your choice, optionally modifies your system path to include the OpenCV binaries, and registers several DirectX filters. By default, it installs to C:/Program Files/OpenCV/<version>.

Customizing a Windows Install

For Windows users, OpenCV is easy to install, and the default installation will work. But a bit of advance planning may leave you happier with the results. Here are a few suggestions.

Since OpenCV is a developers’ toolkit — not a program — you may want to locate it somewhere other than your Program Files directory. If you do prefer to locate it elsewhere, decide that before you run the installer, and enter that location when asked.

I suggest you also decide — before installing — how you want Windows to find the OpenCV dlls. You can either modify your system’s PATH variable to include their location, or you can move them, after installing, from OpenCV’s “bin” directory to your SYSTEM_ROOT directory.

If you prefer to move the dlls, but aren’t sure where your SYSTEM_ROOT directory is, you can locate it by running the sysinfo utility available at www.cognotics.com/utilities.

If you prefer to modify the PATH rather than moving the dlls, you can have the installer do that for you by selecting the check box “Add bin directory to PATH.”

After Installing

The OpenCV directory contains several subdirectories. The docs directory contains html documentation for all the OpenCV functions and datatypes. Since the best documentation is a working example, you might also want to browse the “samples” directory.