Here is a technical report I have written in 2006, which follows an interesting route to increase the steganographic capacity of lossless compressed images such as PNG files. This is done by using a color space which is perceptually more linear than the technical RGB cube: CIELab (or CIELuv).

This is what it looks like in theory:

The point is that in some spots in the right hand colorspace (CIELab), colors as defined in the RGB space are more crunched together than in other places, for example in the blue area. This effect is now exploited by applying a uniform subdivision on the CIELab color space and interchanging indivual, equally perceived colors to encode a message.

Here is the PDF: Using perceptually uniform color spaces for image steganography

Here is a technical report I have written in 2006, which follows an interesting route to increase the steganographic capacity of lossless compressed images such as PNG files. This is done by using a color space which is perceptually more linear than the technical RGB cube: CIELab (or CIELuv). This is what it looks like ...