In steganographic applications emphasis is concentrated on security and on preventing the detection of the hidden information. There are many steganographic techniques that can be found in the literature. Some of these techniques rely on embedding the information in the least significant bits [2]. Other techniques use image palettes [3] to embed information. Still more techniques use spread spectrum modulation techniques [4]. Further details on different steganographic techniques are contained in the following references [5,6,7]. In this paper we describe a new steganographic technique for embedding a large amount of information inside a cover image. The technique is based on the principle of decorrelating the cover image samples. Decorrelating the image samples results in a two dimensional set of independent and identically distributed samples. These samples have a Gaussian distribution in the transform domain. The decorrelated image has components at all frequencies. Therefore, the image transform coefficients will occupy the entire image spectrum, hence producing more coefficients to be used to embed information. Image decorrelation is obtained by using a key which can be either private or public. Data embedding and extraction is performed using quantization principles. The transform coefficients amplitudes are modified in such a way that they resolve whether a binary "1" or "0" is embedded when divided by some threshold . This paper is organized as follows. In Section 2, we describe our data embedding system. In Section 3 we examine the security of the system. In Section 4 we present an example for our proposed system, and Section 5, gives our conclusion and suggestion for future work.