Image watermarking with multi-biometric data

With the widespread use of the Internet, information security becomes more important every day. Because, the information can be easily copied, stolen and imitated and can be used for bad purposes. In particular, many measures are taken to ensure the security of personal information. The most popular of these are personal passwords. However, these passwords are also likely to be stolen. For this reason, biometric features of authorized people can also be used for access to information. Because the biometric data are personal and cannot be imitated. The use of multiple biometric data in combination instead of one increases security. However, these biometric data also need to be stored in a data center. Rather than storing biometric data as they are, storing data after hiding in different data will provide an extra layer of security. For these reasons; it is thought that the biometric data belonging to the person will be placed as watermarks in a host image. Palm print, iris and ear data are used as watermarks, in this thesis. The method used to recover the watermarks is only known by authorized person and it is not possible for unauthorized people to get these watermarks with random methods. While the malicious people try to access this data, they apply different types of attacks to the watermarked image. Depending on the nature of the systems, in some cases it is desirable that the watermarks be completely lost (fragile watermarking) as a result of the attacks, whereas in some cases it is desirable that the watermarks be resistant to attacks (non-fragile watermarking). In this thesis, different methods (singular value decomposition, finite ridgelet transform, contourlet transform) were used to determine suitable methods for non-fragile watermarking and semi-fragile watermarking and scale factors to use in these methods. Iterative and analytical optimization methods were used to determine the scale factors. These methods are spiral optimization, particle swarm optimization and polynomial regression. Keywords: Multi-biometric watermarking, singular value decomposition, finite ridgelete transform, contourlet transform, optimization