DWT DCT Hidden Image Watermarking C#

Implementation of DWT-DCT image watermarking algorithm using c#

Theory and implementation:

The implementation is based on "Combined DWT-DCT digital image watermarking" research paper https://www.researchgate.net/publication/26621646_Combined_DWT-DCT_digital_image_watermarking

Image watermarking Algorithm:

• Apply DWT on target image to split it into 4 sub-bands.
• Divide HL1 sub-band from previous step into 4x4 blocks.
• Take grey-scale watermark image and convert it into vector of 0 and 1 (white and black).
• Generate two uncorelated PN sequences.
• Embed each pixel of watermark image in to mid-band coefficients of DCT blocks using PN sequences.
• Apply inverse DCT to each 4x4 block.
• Apply inverse DWT to produce watermarked image.

Watermark retrieval algorithm:

• Apply DWT on target image to split it into 4 sub-bands.
• Divide HL1 sub-band from previous step into 4x4 blocks.
• Regenerate PN sequences from watermark embedidng step using same seed.
• For each DCT block, find correlation between mid-band coefficients and PN sequences and determine if extracted watermark bit should be 0 or 1.
• Reconstruct the watermark.

Steps are described with a little more details in the paper, but those above should be enough to understand the code if you are familiar with image watermarking.

Results:

Embedded watermark:

Extracted from uncompressed image:

Extracted from 75% jpeg compressed image:

Seems like the result is not very robust to jpeg compression. In addition, the original image gets quite visibly distorted if strength of embedding is increased.