Computers have invaded all premises of the human world, starting from a grocery store to a missile launching center. Because of the omnipresence of computers, it becomes more and more difficult everyday to secure the confidential information from misuse. The fairly common technique of cryptography has been proved inadequate in recent years. Steganography, a contemporary yet an age-old technique to hide secret data into an unsuspected cover media like an image, thereby preventing the recognition of the very presence of secret data, is an alternative. In this study, an improved image steganographic approach is proposed. This method reduces the mean square error (MSE) by localizing the error-reduction process to every row. The error reduction is performed by selective embedding of the actual secret, its binary complement, gray-coded version or inverted gray-coded version. Of the four versions, the version giving the least MSE is embedded on a row-by-row basis. This method reduces the MSE by a factor of 1.8 and boosts the peak signal to noise ratio (PSNR) by a 0.25 db and considerably increases the security.