Hi,
For properly hiding a secret image under a cover image(steganography), the preprocessing steps are also equally important, where we convert the images from “rgb2gray” and resize the images secret image to fit in the cover images. I am assuming you have done these steps as well. DCT coefficients represent the frequency components of an image block. Generally, the lower frequency components contain more notable image information, while higher frequency components contain finer details.
Quantization disproportionately affects higher frequency components by using larger divisors for these coefficients. This results in many of these coefficients being rounded to zero or a very small value. When quantization is applied, the subtle modifications made to these higher frequency coefficients for embedding the secret data can be lost due to rounding. Therefore, you can skip the quantization step here.
You can refer to the following documentation to know more about dct.
