Semi-fragile image watermarking method based on index constrained vector quantisation Zhe-Ming Lu, Chun-He Liu, Dian-Guo Xu, Sheng-He Sun, IEE Electronics Letters , Volume: 39 Issue: 1 , Jan. 2003 Page(s): 35 -36 Advisor :Dr. Chang, Chin-Chen Reporter :Jian, Yong-Xin Date :2003/6/10
Outline INTRODUCTION PROPOSED METHOD EXPERIMENTAL RESULTS CONCLUSIONS
INTRODUCTION robust Semi-fragile fragile VQ DCT DWT
PROPOSED METHOD Original image "Lena“ (gray scale, 512 × 512) Block size=4 x 4 Divided into 16384 blocks (512x512)/(4x4)=16384 Watermark W (binary scale, 32 × 32) To be embedded 1024 bits
PROPOSED METHOD (cont.) Watermarking position & bits input vectors index constrained VQ encoder division encoded indices watermarked output vectors normal VQ decoder composition image EMBEDDING WATERMARK
PROPOSED METHOD (cont.) Step1.Use LBG to generate a codebook. Step2.Divide image and find its best codeword. codebook index 1 2 . 225 4093 4094 4095 225 original image
PROPOSED METHOD (cont.) Step3.Embed bits by order. Step4.Replicate 16 times. 225: 000011100001 1 12bit/index 225 Step5.Save the position information watermark
EXTRACTION WATERMARK input vectors normal VQ encoder division encoded indices watermarking position extracted watermark extracted bits composition check bit EXTRACTION WATERMARK
EXPERIMENTAL RESULTS Original Lena image Watermarked Lena image PSNR=31.85 db
EXPERIMENTAL RESULTS (cont.) Original watermark No attack, NC=1.0 JPEG, QF=100%, NC=0.9926 JPEG, QF=80%, NC=0.8053 VQ1, N=4096, NC=1.0 f. VQ2, N=8192, NC=0.8367 g. Image cropping, NC=0.9009 h. Median filtering, NC=0.0640 i. Blurring, NC=0.1233 j. Rotation, NC=0.0556
CONCLUSIONS A semi-fragile digital image watermarking method is presented based on index constrained VQ. The method is simple and efficient. The proposed algorithm can be applied to the case of multiple watermarking by select more watermarking positions.