1 %STC_PM2_DLS_EMBED Embeds message in a cover object with +-2 changes - DLS.
3 %[D Y NUM_MSG_BITS L] = STC_PM2_DLS_EMBED(X, COSTS, MSG, TARG_D)
4 % implements Distortion Limited Sender. Embeds as many bits as possible
5 % into cover object X, such that the cost of stego object Y is around
6 % TARG_D with changes limited to +-2. Uses default constraint height
7 % h=10, expected coding loss 5% and INF for cost of wet elements. Outputs
8 % distortion caused by embedding D, stego object Y, number of message
9 % bits in each layer of LSBs NUM_MSG_BITS and coding loss L. See [1] for
12 %[D Y NUM_MSG_BITS L] = STC_PM2_DLS_EMBED(X, COSTS, MSG, TARG_D, EXP_L)
13 % the same as above, but uses expected coding loss EXP_L - see [1].
15 %[D Y NUM_MSG_BITS L] = STC_PM2_DLS_EMBED(X, COSTS, MSG, TARG_D, EXP_L, H)
16 % the same as above, but uses STCs with constraint height H - see [1].
18 %[D Y NUM_MSG_BITS L] = STC_PM2_DLS_EMBED(X, COSTS, MSG, TARG_D, EXP_L, H, W)
19 % the same as above, but uses W as the cost of wet elements - see [1].
21 % COSTS array must be 5xN, where N=numel(X).
22 % COSTS(1,i) = cost of changing X(i) by -2
23 % COSTS(2,i) = cost of changing X(i) by -1
24 % COSTS(3,i) = cost of changing X(i) by 0
25 % COSTS(4,i) = cost of changing X(i) by +1
26 % COSTS(5,i) = cost of changing X(i) by +2
28 % NUM_MSG_BITS contains number of bits embedded in different layers
29 % NUM_MSG_BITS(end) - # of bits in LSBs
30 % NUM_MSG_BITS(end-1) - # of bits in 2LSBs
31 % sum(NUM_MSG_BITS) = total number of embedded bits. This can be less
34 % Use STC_ML_EXTRACT(Y, NUM_MSG_BITS, H) to extract the message back.
36 % Author: Tomas Filler email: tomas.filler@gmail.com
37 % www: http://dde.binghamton.edu/filler
39 % STC Toolbox website: http://dde.binghamton.edu/filler/stc
42 % [1] T. Filler, J. Judas, J. Fridrich, "Minimizing Additive Distortion in
43 % Steganography using Syndrome-Trellis Codes", submitted to IEEE
44 % Transactions on Information Forensics and Security, 2010.
45 % http://dde.binghamton.edu/filler/pdf/Fill10tifs-stc.pdf
47 % [2] T. Filler, J. Judas, J. Fridrich, "Minimizing Embedding Impact in
48 % Steganography using Trellis-Coded Quantization", Proc. SPIE,
49 % Electronic Imaging, Media Forensics and Security XII, San Jose, CA,
50 % January 18-20, 2010.
51 % http://dde.binghamton.edu/filler/pdf/Fill10spie-syndrome-trellis-codes.pdf
53 % [3] T. Filler, J. Fridrich, "Minimizing Additive Distortion Functions
54 % With Non-Binary Embedding Operation in Steganography", 2nd IEEE
55 % Workshop on Information Forensics and Security, December 2010.
56 % http://dde.binghamton.edu/filler/pdf/Fill10wifs-multi-layer-stc.pdf
58 % See also STC_ML_EXTRACT, STC_PM1_PLS_EMBED, STC_PM1_DLS_EMBED
59 % STC_PM2_PLS_EMBED, STC_EMBED, STC_EXTRACT.