부산대학교 도서관

In prediction-error-expansion and pixel sorting based reversible data hiding (RDH) works, only part of the neighboring pixels are used in pixel prediction and local complexity calculation. To fully exploit the local correlations, we propose a novel efficient RDH using filter selection. Firstly, we divide the cover image into multiple non-overlapping layers following the concept of image down-sampling. Then data is embedded in a layer-wise manner. For data embedding, the prediction-error and the local complexity are estimated using the idea of block-based image filtering, the related techniques of which have been widely studied in the area of image processing. Experimental results validate the applicability of the proposed work in the perspectives of both embedding capacity and image qualitv.