Watermarking Method Using Concatenated Code for Scaling and Rotation Attacks

Abstract

Abstract. We proposed a watermarking method using a concatenated code and evaluated the method on the basis of IHC evaluation criteria. The criteria include JPEG compression, clipping, scaling, and rotation as attacks. For the robustness of messages, we introduced concatenated code, since it has a high error corrective ability to decode messages against JPEG compression. When a region is cropped from a stego-image, the position of watermarks might be unclear. Therefore, markers or synchronization codes were embedded into the stego-image. Since scaling causes pixel loss, and rotation causes distortion, watermarks were embedded into minified images. Quantization index modulation was used for embedding and extracting the watermarks without the original images. As a result, our method was evaluated on the basis of highest image quality and could achieve an average peak signal-to-noise ratio of 36.250 dB. Moreover, our method was evaluated on the basis of highest tolerance and could achieve an average compression ratio of 2.633% without errors.

Image Watermarking Method Satisfying IHC by Using PEG LDPC Code

“…4. More precisely, the proposed method is able to extract a watermark at the same computational cost as that of Hirata and Kawamura's method [13] and yields stego-images of higher quality than those obtained using the method proposed by Ogawa et al [14] under the condition of IHC ver. 4.…”

“…In other words, the attacks may cause errors in the watermark. Digital watermarking methods avoid these errors by using spread-spectrum techniques or correct them by using error-correcting codes [4]- [6], [10]- [13]. A watermarking method that uses an error-correcting code embeds as a watermark the codeword generated from a message.…”

“…4, one proposed by Ogawa et al [14] and the other proposed by Hirata and Kawamura [13]. The former encodes the mesCopyright c 2017 The Institute of Electronics, Information and Communication Engineers sages by using a convolutional code and embeds the watermarks in 256 × 256-pixel blocks in the 2D DCT domain.…”

“…The method proposed by Hirata and Kawamura [13] encodes the messages by using a concatenated code [19], [20] combining BCH and low-density parity-check (LDPC) codes, and it embeds the watermarks in 8 × 8 blocks (i.e., 8 pixels by 8 pixels) in the 2D DCT domain. The synchronization code used for synchronization recovery is also embedded in the image.…”

“…(1) A marker is embedded in small blocks for extraction as fast as that in the method proposed by Hirata and Kawamura [13]. (2) Tolerance against scaling and rotation is obtained by increasing of number of the same watermarks.…”