Steganographic algorithms in the recent past have been producing stego images with perceptual invisibility, better secrecy and certain robustness against attacks like cropping, filtering etc. Recovering a good quality secret from a good quality stego image may not always be possible. The method proposed in this paper works in transform domain and attempts to extract the secret almost as same as the embedded one maintaining minimal changes to the cover image by using techniques like median maintenance, offset and quantization.