Steganography based on neighbourhood pixels

This paper proposes a steganography technique to hide secret messages in the least significant bits of the pixels of the cover image which are most different from their neighbourhood pixels. Since such pixels are the areas in an image which correspond to great changes in the visual property, they make a better option to hold secret data than smoother regions of the image, where a small distortion is much more noticeable. This paper uses LSB Matching algorithm for data embedding and the embedding method is adaptive to the amount of data in the message. Smaller is the amount of data, higher the value of threshold is chosen. The proposed technique is tested on BOWS2 Database consisting 10000 images and is compared with current steganography techniques for their robustness against various steganalysis methods. The proposed technique has given overall better results as compared to other techniques.