Compression resilient image watermarking scheme in spatial domain

II. REGION SELECTION BASED ON ENTROPY VALUES Abstract—The present paper describes a compression resilient data hiding scheme in spatial domain by inserting data in different blocks in the cover image. The blocks are selected based on the value of information content that is measured using an exponential form of entropy. Experimental results show the resiliency of the scheme against several unintentional and deliberate attacks such as lossy compression, linear and non linear filtering, cropping, noise addition, image sharpening etc. Low computation cost, less visual impairments and blind detection is the novelty of the scheme. The present watermarking scheme is a block based spatial domain approach where embedding regions are selected on the basis of average information or entropy of pixel intensities. Entropy values are used because information about an image is captured not from the single pixel alone but from the pixel and its neighborhood, since a strong two dimensional (2D) spatial correlation exits among the neighboring pixels in most of the natural images. If data-embedding process creates a noticeable change in spatial correlation among the neighboring pixels, image visual quality may be degraded and the imperceptibility requirement of watermarking will not be fulfilled. One of the good measures of spatial correlation among the neighboring pixels is the average information value. In Shannon’s entropy measure [4], the signal is considered as a long sequence of symbols and entropy value depends on the relative occurrence of symbols irrespective of their position of occurrence. But digital image being a two dimensional sequence of highly correlated pixel values, average information or entropy of image (or sub image) not only depends on the relative occurrence of the pixel values but also depends on their position of occurrence. It has been shown in [5] that exponential form of entropy function can capture pictorial information better compared to conventional Shannonian entropy. The authors of [5] defined the entropy of an image block n=(a x a) in the form Index Terms Image watermarking, entropy, lossy compression and negative modulation.