A high payload data hiding scheme based on modified AMBTC technique Multimedia Tools and Applications : SCI 1.5 in 2018 Aruna Malik :来自印度勒克瑙北方邦技术大学和M.Tech的计算机科学与工程专业。 Geeta Sikka目前在贾朗达尔国立技术学院计算机科学与工程系担任副教授。 Harsh K. Verma是印度贾朗达尔国立技术学院计算机中心的教授和主任 Sourse: Multimedia Tools and Applications, Volume 76, Issue 12, June 2017, pp 14151–14167 Authors: Aruna Malik, Geeta Sikka, Harsh K. Verma Speaker: Guan-Long Li Date: 2018/8/23

Outline Related Work Proposed method Experiment Results Conclusion Absolute moment block truncation coding(AMBTC) Proposed method Experiment Results Conclusion

Related Works - AMBTC k: n × n image block x: image block ( 𝑥 1 , 𝑥 2 ,.., 𝑥 𝑘 ) q: the number of pixels greater than the threshold value Cut grayscale images into non-overlapping image blocks One bit map BM and two quantization levels a , b to represent the block Encoded as {a, b, B} after compression Step 1 : Calculate the mean value 𝑥 of the block Step 2 : Create a bit plane : the original image pixel value < 𝑥 ,‘0’ in the bit map. the original image pixel value > 𝑥 ,‘1’ in the bit map. Step 3 : Calculate the quantization levels : Calculate the momentum absolute value 𝛼 Calculate the low mean value as quantization level a of the block Calculate the high mean value as quantization level b of the block 將灰階影像切割成不重疊的影像區塊 一個位元圖BM及兩個重建階 a , b 來表示該區塊 壓縮後編碼為(a, b, BM) Lema M, Mitchell O (1984) Absolute moment block truncation coding and its application to color images. IEEE Trans Commun 32:1148–1157

Related Works - AMBTC quantization level a: 8 9 8 6 7 9 10 12 11 8 13 14 6 7 1 Mean Value : 10 10 12 11 13 14 quantization level b: 12

Proposed method Input- I: original image of size N × N pixels, thr: threshold, S: secret data bit stream. Output- AMBTC compressed stego codes Step 1: Divide input image I into 4 × 4 non overlapping blocks in raster scan order. Step 2: Process each block using AMBTC scheme to get two quantization levels a and b, mean value, and a bit plane B. Step 3: Reprocess each block to get two bit plane B’ so that caused noise due to the compression is reduced. • the original image pixel value < a ,‘00’ in the bit plane. • a <= the original image pixel <= mean value, ‘01’ in the bit plane. • mean < the original image pixel value < b, ‘10’ in the bit plane. • b < the original image pixel value ,‘11’ in the bit plane. Step 4: Calculate absolute difference value D for the block , such that D = |a − b|.

Proposed method Output : {a’, b’, c’, d’, B’} |a’-b’|<=thr Calculate four new quantization levels Replace bits of the two bit plane with 32 bits Output : {a, b, c’, d’, B’} D<=thr |a’-b’|>thr Step 4 |a’-b’|<=thr Output : {a’, b’, c’, d’, B’} D>thr Replace first LSB of the two bit plane with 16 bits Calculate four new quantization levels Output : {a, b, c’, d’, B’} |a’-b’|>thr Four new quantization levels 𝑎 ′ = 1 𝑞 00 𝑥∈ 𝐺 00 𝑥 , 𝑏 ′ = 1 𝑞 11 𝑥∈ 𝐺 11 𝑥 𝑑 ′ = 1 𝑞 01 𝑥∈ 𝐺 01 𝑥 , 𝑐 ′ = 1 𝑞 10 𝑥∈ 𝐺 10 𝑥

Proposed method 100 90 85 82 76 75 72 77 80 81 79 1 11 10 00 01 AMBTC Mean = 81 Original block One bit plane a=77, b=85 Two bit plane

Proposed method 11 10 00 01 10 01 00 11 Embed in Smooth block Two bit plane Stego block a’=78, b’=78 c’=83, d’=81 Secret = (10100101001011100010111011100101)

Proposed method 186 211 212 205 181 208 202 185 203 193 194 190 1 00 11 10 01 AMBTC Mean = 200 Original block One bit plane a=188, b=207 Two bit plane

Proposed method 00 11 10 01 01 11 10 00 Embed in complex block Two bit plane Stego block a’=187, b’=207 c’=207, d’=190 Secret = (1110010100001010)

Proposed method 10 01 00 11 D = |a’-b’| = |78-78| = 0 < thr {a’, b’, c’, d’} ={78,78,83,81} Secret = (10100101001011100010111011100101) 01 11 10 00 D = |a’-b’| = |187-207| = 20 > thr {a’, b’, c’, d’} ={187,207,207,190} Secret = (1110010100001010)

Experiment Results

Experiment Results Ou et al. :提出了一種基於絕對矩塊截斷編碼（AMBTC）的改進圖像隱寫方案。所提出的方案的目的是同時實現高有效載荷，良好視覺質量和低計算複雜度。在該方案中，預定義閾值以將AMBTC壓縮代碼的塊分類為平滑或複雜塊，然後嵌入數據。對於平滑塊，它們的位平面用於嵌入數據。之後，重新計算平滑塊中的兩個量化級別以最小化圖像質量的失真。對於復數塊，通過交換兩個量化級的順序一起切換比特平面來隱藏一部分秘密比特，通過該比特平面可以增加有效載荷而沒有任何失真。此外，所提出的方案繼承了AMBTC方法的優點，例如令人滿意的圖像質量，易於實現和低計算複雜度。通過可調閾值，所提出的方案的應用變得靈活，這意味著不同的閾值可以用於不同的應用。實驗結果和分析證明了該方案的有效性和優越性。 [20]Ou D, Sun W (2014) High payload image steganography with minimum distortion based on absolute moment block truncation coding. Multimed Tools Appl 74(21):9117–9139 [21]Yang B, Schmucker M, Funk W, Brush C, Sun S (2011) Integer DCT-based reversible watermarking for images using companding technique. Proc Int J Electron Commun 65:814–826 [18] Petitcolas F, Anderson R, Kuhn M (1999) Information hiding – a survey^. Proc IEEE 87(7):1062–1068 [22] Zhang Y, Guo S, Lu Z, Luo H (2013) Reversible data hiding for BTC-compressed images based on lossless coding of mean tables. IEICE Trans Commun 96(2):624–631

Conclusions This scheme embeds almost 2 times more secret data than the existing schemes with better visual quality. The scheme affects the AMBTC compression ratio.