A Steganographic Algorithm Based on Image SparseDecomposition Optimized by GA

Considering the sparsity and integrity of the spare representation of images over over-complete dictionaries, this paper presents a novel image steganographic method with genetic algorithm (GA) based on spare decomposition. In this method, the data hiding process is integrated into the image spare compression process. First, in each iteration of the matching pursuit of image spare decomposition, the best matching atom is selected by GA. Then, the coefficients of sparse decomposition are quantified by different quantization bits. Finally, the stego image is obtained via embedding secret information in the different least significant bits (LSBs) of the quantized coefficients. Experimental results show that the proposed steganographic algorithm maintains good invisibility. Meanwhile, compared to the classical LSB methods of spare domain and DCT domain, the new steganography has better ability in resisting steganalysis under the same embedding capacity. Experimental results also indicate that the new steganography is less sensitive to the number of the embedding bits, leading to good expandability in embedding capacity.