I want to make a DCT steganography on the quantized DCT coefficients... should I do the embedding block by block?? and can I use the blockproc function in Matlab for embedding ?

DCT steganography embedding function using Matlab

Answer 1

[Walter Roberson]

Any block that you are not going to embed data in, you might as well not bother to put through the DCT process.

DCT, like FFT, is not specific to any one block size. However, DCT for information hiding is often used together with JPEG images. JPEG uses DCT on 8 x 8 blocks internally, so if you do DCT on 8 x 8 blocks yourself and hide data in them, then you know that the data hiding would survive the JPEG DCT process, whereas if you used a different block size than JPEG was using, if you were to then compress the image using JPEG, the hidden data might get lost.

[Budoor Salem]

Thank Mr.Walter for your answer... yes I am doing a steganography on JPEG images I made the 8X8 blocks and the DCT..now I am on the embedding stage.. what do you suggest for the embedding..should I do the embedding block by block...can I use the blockproc function through the embedding or what do you suggest please and thank you

[Walter Roberson]

It does not really matter whether you use blockproc to embed the data or not. But remember that if you do use blockproc, then the function you invoked on each block would have to figure out which of the bits were to be embedded in that particular block. That would probably involve looking at the "location" field of the block_struct passed in to the function.

You might find it easier to blockproc to do the dct, then use normal array indexing to store bits in the whole array of coefficients, then use blockproc to do the inverse dct.

[Budoor Salem]

really thank you for your answer..this is really help...so thanks... now just from what I understood now..is that the blockptroc function I will use it in the DCT and inverse DCT...whereas in the embedding ..I embed in the whole DCT coefficients of the image as a whole.not by block by block...as if I am doing LSB steganography on the whole image without dividing it into blocks?? am I correct?? I think it would be easier this way..

but is it like I am supposed not to embed in the every DC coefficient of each block (the first coefficient of each block) ,,then how will I keep track of the DC coefficient ..

thank you again I really appreciate your valuable time answering me..your answer really helped me figuring out what I will do Thank you

[Walter Roberson]

After using blockproc on YourImage to create dct_coefficients,

num_blocks = floor( size(YourImage)/8 );
tot_blocks = min( prod(num_blocks), number_of_bits_to_embed);

for blockidx = 1 : tot_blocks
  [rblock_idx, cblock_idx] = ind2sub(num_blocks, blockidx);
  first_dct_coeff_of_block = dct_coefficiants((cblock_idx-1)*8+1, (rblock_idx-1)*8+1);
  %now modify the coefficient and store it back in the same place
  ...
end

Note: the range of values that can be in first_dct_coeff_of_block is +/- 255*8*8 which is +/- 16320 . The values are floating point numbers, not integers.

[Budoor Salem]

God bless you Mr. Walter.. Thank you so much I appreciate your help :D