High Capacity Data Hiding Techniques for Digital Images

Abstract

Today, the demand of internet has made the transmission of digital media much easier and faster. Open nature of internet, risks of illegitimate accessing and unauthorized tempering with transmitted data is increased day by day. Protection of secret information from unauthorized users in a public network has become an important issue. Data hiding is one of the most demanding techniques to protect the security of digital media. In this dissertation, we have proposed three data hiding techniques. We have proposed a histogram based reversible data hiding technique for digital images. In this technique, cover image is divided into blocks of equal size. Then modulus operator is used to increase the occurrence of peak points in the histogram of blocks. Extracted secret image is similar to original secret image. Maximum capacity is more than 96,000 bits and Peak Signal Noise Ratio (PSNR) is also higher than 50 dB. Embedding capacity and PSNR are higher than existing histogram based data hiding techniques. Histogram based reversible data hiding technique is modified to embed more secret data using the concept of multilevel embedding. Modulus operator is also used to increase the occurrence of peak points of histogram of blocks. Secret information is embedded into peak points of histogram but in the modified technique minimum point is not necessary to a zero point. Extracted secret image and cover image are similar to original secret image and original cover image respectively. Maximum capacity is more than 2.6 bit per pixel (bpp) and PSNR is also higher than 44 dB. Proposed technique provides better embedding capacity and better visual quality of marked image than existing multilevel data hiding techniques. Interpolation based blind data hiding technique is also proposed in this dissertation. In this technique, difference images are generated using interpolation of sub sampled images and on the basis of even or odd state of the difference pixel value; the secret information is embedded into the cover image. Maximum capacity is more than 196,000 bits and PSNR is also higher than 50 dB. Proposed technique provides higher embedding capacity, better visual quality marked images and less computational complexity than interpolation based data hiding techniques