I have an idea of applying DCT on an image and then embed text into the LSB of the coefficient of DCT and transmit the image to B (the receiver). Now, using the same procedure, can B transmit another watermark along with the previous one with this method to another person? Is this a suitable procedure?
Since you usually compress images, which would do something very similar, at least in the case of jpeg compression, but would completely omit or severely reduce LSBs, I guess, no, in practice not.
If you're sending around full-color pictures uncompressed, at least I'd look at the entropy of your LSBs and the PSD, and the latter would probably lead to the suspicion that you're using steganography.