The popularity of social networking and picture sharing websites continue to increase, so developers of these websites try to add more and more functions to these pages. Naturally, this trend affects the services connected with pictures as well. Nowadays one of the most popular features are face detection and face recognition. With this aid, we can easily categorize our pictures, and search for friends. But the opportunities in face detection and in face recognition are not limited to the functions mentioned above, they can be also used to verify the users identity, as seen on many websites.
In my thesis the scheming and the implementation of a face detection system is examined. In the first chapter, a brief history overview of digital image processing can be read from the 1920s till nowadays. Then in chapter 1.3, a short list of different type of face detection systems can be found with the detailed overview of the Viola-Jones detector.
In the next part my own digital processing framework is designed, then implemented based on pipeline architecture. In this section a detailed description of the modules can be read. During the designing part, there were three important requirement to comply with : performance, accuracy and robustness. Considered these aspects, C++ and OpenCV were chosen to be used.
In chapter three the implementation procedure is described including the mathematical background for each module.
After implementing the algorithm, the next phase was to test it. FEI face database - with 200 individuals - was used to measure the performance and accuracy of the application. It detected all of the faces, with two positive false detection only.
Finally, the future potentials are discussed, including a short introducing to a face recognition system.