Developing a block-wise interpolation based adaptive data embedding scheme

Data hiding schemes based on quadratic interpolation have the potential to offer better image quality and higher embedding capacity. Being reversible, such schemes also allow user to restore the embedded image back to its original version, which is the key requirement of many applications like medical and military imaging. This paper presents a simplified data hiding scheme based on quadratic interpolation. Particularly, we apply quadratic interpolation in non-overlapping blocks of an input image. Instead of calculating multiple interpolated values considering neighborhood pixels in multiple directions, we simply consider only one (i.e., either horizontal, vertical or diagonal) direction for interpolation and correlate the interpolated values both with and without corresponding local mean. This significantly reduces the complexity and offers high embedding capacity as high as 2.5 Mbits with better or similar image quality, while compared with a recently published interpolation based scheme.