This is a continuation of my blog post on intro to digital watermarking. In that post, we discussed what digital watermarking is and how it can be achieved. Here, we will discuss the intuition behind image watermarking and a few techniques that can be used. If you look at enough number of images, you will realize that not all of them are equally suited for watermarking. At least, we cannot use the same criteria to watermark all the images. How do we know where to watermark an image? Are there any rules or do we just place some watermark randomly in an image? Does it make a difference?
Here are the things we to need to consider before watermarking an image:
- A busy image hides a watermark well, you will be able to easily apply a strong watermark signal without impacting your image quality. This is one of the core concepts of image steganography. Busy images help us in hiding the watermarks because the human visual system is not sensitive to high frequency regions in an image (high frequency regions correspond to busy areas like textured surfaces, regions with lot of leaves and trees, many small objects placed in a small area etc). If you look at the image in the first paragraph carefully, you will notice a couple of hidden watermarks. You can see the words “Hidden Watermark” immediately but the other watermarks are not immediately apparent because it’s a busy image.
- Areas of bright highlights or dark shadows do not accept watermarks well. In highlights, there are too few pixels to modify to add the signal. In shadows, the ability to modify pixels is essentially halved because of saturation. Basically, these constitute low frequency regions, which our visual system is sensitive to. Even if a surface is textured, it becomes a low frequency region when highlighted.
- For the same reason that highlights do not accept watermarks well, pure white areas cannot be watermarked at all. If you must watermark a pure white area, we should add a light color tint into larger white areas. It will need to be dark enough to reproduce for sufficient color coverage to embed a detectable watermark. Unfortunately, the watermark may become more visible with the added detectability.
- In order to identify areas for different watermark strengths, we can use the same criteria we normally use for unsharp masking. For example, faces will probably require a lower watermarking strength; areas with lots of detail can take a watermark of higher strength.
————————————————————————————————-
Perpetual Enigma 