A skin tone detection algorithm for an adaptive approach to steganography

Digital image steganography: Survey and analysis of current methods

“…The majority of steganography research to date has overlooked the fact that utilising objects within images can strengthen the embedding robustness -with few exceptions. A steganography approach reported in [92,93], incorporated computer vision to track and segment skin regions for embedding under the assumption that skin tone colour provides better embedding imperceptibility. They used computer vision techniques to introduce their rotation and translation invariance embedding scheme to establish an object oriented embedding (OOE).…”

“…23. Resistance to lossy compression thanks to the DWT Performs better than DCT algorithms in keeping the carrier distortion to the minimum Ability to embed secret data into different orientation, acts as an additional secret key Re-orienting the stego-image to its origin will invoke interpolation, thus providing a mask that fools any statistical attack [93]. Embedding into the 'Y' channel has the advantage of better resistance to compression, while embedding into 'Cr' channel has the advantage of better image perceptibility at the expense of resistance to image compression.…”

Efficient Skin Detection under Severe Illumination Changes and Shadows

“…The methods considered are the skin locus approach in normalized RGB (NRGB) [18], the use of hue and saturation components in HSV color space (HS) [8], the combination of normalized RGB with hue and saturation (NRGB-HS) [19], the use of chromatic components in YCbCr color space (CbCr) [7], and the luminance based method described by Cheddad et al [4]. For all these methods we use the thresholds given by authors.…”

“…However, it was also shown that discarding the luminance component degrades skin color segmentation results. For example, Cheddad et al [4] propose a luminance based skin detection method based on deriving an error signal from the grayscale map and the non-red encoded grayscale version of the image. A comprehensive survey of skin color detection in different color spaces can be found in [1,2].…”

“…In most approaches, a color space that de-correlate luminance and chromatic components is adopted, since it is assumed that chromatic components are less sensitive to lighting conditions [3]. However, Cheddad et al [4] have shown that luminance can be useful in finding skin pixels in color images.…”