Maximum of image watermarking scheme has been implemented in DWT domain. In that also watermarking is done in LL Sub band . Why watermarking is not preferred in other sub-band specially in HH sub-band.
The high capacity is one motive, coefficients are not negative and can be treated as integer values. There are some attempts to use all bands, for the non-LL, researchers track down the captured edge areas.
Dear Sanjay
The DWT decomposed image in LL and other high frequency sub bands (HL, LH, and HH). The most of the energy is laid in LL band. The other bands are having detailed information and less energy. In watermark application, the watermark is embedded in LL band because it will not much effect the energy or change coefficient value so its effect is not notice able after reconstruction. average information. The other bands coefficients are vary small and adding the watermark will effect the detailed information and effect is visible in reconstructed image.
The researcher has adopted a digital watermarking technique which operates in the frequency domain: a hybrid watermarking scheme based joint discrete wavelet transform – discrete cosine transform – (DWT-DCT).
Generally, DCT and DWT are initially performed and then transform the image into the spatial domain. Grayscale images are mostly preferred in watermarking schemes. One of the advantages in DCT-based algorithms is its high capacity of hiding data. Though these algorithms withstand various attacks (robust), it is also tamper-prone and can easily deteriorate the quality of the watermarked image. DWT-based algorithms have better image quality but are vulnerable to various attacks. To solve these problems, a hybrid digital watermarking scheme basing on DWT-DCT is used.
Refer the following link.
https://pdfs.semanticscholar.org/5394/1287b08b892239cdc5c6b2e0bb9479061bc1.pdf
most of the energy of the image is presented in LL Sub band , even we can do inverse DWT and recover the image only by using LL Sub band . Also this sub-band provides more robustness in compare with other sub-bands...that's why most researchers prefer this sub-band for embedding.
most of energy in the LL then get robustness but loss imperceptible so many researcher in the last year prefer HL or LH to get robustness with imperceptible
because if you are using HH get imperceptible and loss robustness