A computation structure for 2-D DCT watermarking

In this paper, a new computation structure of image watermarking in DCT domain is presented. This structure employs a single DCT block, instead of two in many existing DCT watermarking systems, for the DCT computation of the image signal and that of the watermark. Using the single DCT block does not double the time required for the two 2-D DCT computations, as the amount of the calculations in this structure is effectively reduced compared to those with two DCT blocks. By using a special condensed 2-D DCT algorithm, the DCT computation of the watermark is divided into two parts and only the part necessary and sufficient for the watermarking is performed to achieve a significant reduction of the overall computation. This structure can be easily implemented into a circuit. Using the single DCT block, the hardware reduction in the 2-D DCT part of the watermarking circuit can reach 50% with the time delay increase of only 34%. It has been confirmed by the preliminary results of the FPGA implementation.

1. Introduction

Digital watermarking is an important mean for identification and authentication of original data such as images or video [1]. It is usually done by embedding a watermark signal into the original signal and this embedding is usually performed in a transformed domain. The 2-D DCT is widely used in image watermarking, in particular, the invisible watermarking [1].