A Block-Based DCT Image Steganography using L-estimation approach and Sparse Reconstruction

Abstract

The paper proposes a transform-domain image steganography approach that is based on the discrete cosine transform and sparse reconstruction principles. The proposed approach embeds secret information in the form of sinusoidal components within a limited number of randomly selected image blocks. This allows controlled and imperceptible data embedding. To improve robustness against non-stationary image content, L-estimation filtering is employed. The goal of the filtering is to isolate stationary frequency components in the discrete cosine transform domain. The remaining coefficients are subject to the compressive sensing reconstruction algorithms in order to recover embedded signal from a reduced set of coefficients. Experimental results demonstrate that the proposed approach achieves a favorable balance between embedding capacity, imperceptibility, and robustness. The combination of random block selection, sinusoidal embedding, and sparse reconstruction further improves security and reduces detectability, making the method suitable for secret communication in digital image applications.