Dithering manipulation chooses the different patterns used to dither images to encode a message. It can achieve a high data rate, 1 hidden_byte/4 cover_bytes. Its major drawback is to rely only on dithered images and to be extremely sensitive to errors in the stego-image.
This approach tries to embed data into regions that are masked by others with respect to the Human Visual System. The embedding can be done in the spatial or in the frequency space. The amount of data that can be hidden depends on the cover image and usually the choice of the areas to be modified has to be done under human control.
These techniques are used every time a lossy compression algorithm is on the signal path. Since those algorithms usually process the image using a DCT (Discrete Cosine Transform), which produces a matrix of floating point coefficients, while the same coefficients will be stored as ints, a rounding process is involved. The embedding of data is acquired piloting the rounding function.
These methods are sometimes considered as variants to the LSB method.
[WW98]
[MBR98]
Previous: Orthogonal projection coefficients manipulation
Next: Moving images
Matteo Fortini