Personal data protection in digital images

OData support
Dr. Kovács Gábor
Department of Telecommunications and Media Informatics

The Internet as a part of modern life, there are many advantages and more functions for us. One example is social media sites, which function as an alternative for basic communication between peoples. So to speak, those are the fastest way to communicate and inform our friends with incredible amount of information. Today's world of rapidly evolving technology, it is available to more people, and to use the full range of all types of Internet applications, such as social media services.

Due to the increasing amount of data, obeying the personal privacy law and its difficulties are more concerned. As one possible solution to this, I have designed a webservice that detects the faces on the picture and makes them unrecognizable.

In the first part of my thesis I studied the implementation of the personal privacy and its shortcomings, particularly in visual data. In addition, I discussed some of the problems which occur on the social networking sites. In some points of view, the webservice that I designed may be a solution to these problems.

In the second part, the Viola-Jones face detecting algorithm is presented in detail, which provides the basis of the web-based service. I chose this algorithm because of its high detection rate with a relatively fast image processing time.

The third part of my thesis is a design process of the webservice, which contains a description of the components and the communication protocols used.

In the fourth part, I achieved the face detection using C + + programming language with the help of the OpenCV library. The detector has been tested with various images collected from the World Wide Web, which include mostly frontal-oriented faces. In order to analyze the detection errors I used test sets which compile images with different head orientations.

Finally I’ve done blurring of the detected faces that results corresponded to my expectations.


Please sign in to download the files of this thesis.