A New Pixel Value Decomposition based Image Steganography Method

Abstract

In this paper, we present a high quality data hiding method using pixel value decomposition that generates higher number of bit-planes than that of the native number representation. Battisti et. al. discussed a data hiding method based on the Fibonacci number bit-plane decomposition. This scheme was further improved by Dey et. al. by using the prime and natural number decompositions. The approach based on natural number decomposition generates 23 bitplanes, requiring a look-up table of large size and the approach based on prime number decomposition generates 15 bit-planes. The weights assigned to the bit-planes increases according to prime numbers and natural numbers. Having large weight values creates large variations in pixel values after embedding the secret data that will result in higher image distortion. In our proposed method, we use a novel set of weights of decomposition, by carefully selecting the weights of decomposition that generates 10 bit-planes. The weights of the first five LSB bits increase slowly and the remaining, grow exponentially. By doing this, the variation in the pixel values after embedding the secret data is quite small and also, due to less number of bit planes the size of lookup table is considerably reduced. As a result, our proposed method maintains less distortion of the stego-image and at the same time has smaller size of the lookup table.