This project aims to propose a method for improve the embedding efficiency of a grayscale cover image for image steganography. Linear coding (matrix embedding) method is used for embedding the message in the cover image so that the capacity of the cover image is utilised efficiently. Compressed domain embedding techniques has proven to be resistant against many steganalysis techniques, yet the capacity of the medium is constraint. The number of non-zero quantized coefficients will be less than the actual number of DCT coefficients, which reduces the capacity of the cover image. This project proposes a method to embed the message using an extension to hamming code, which was proven to improve the embedding capacity. It also compares the results with existing techniques.