In this thesis, the high and low reconstruction levels of each block must be obtained first by using block truncation coding. Then, the distance value of each block can be calculated by subtracting its low reconstruction level from its high reconstruction level. Meanwhile, the predicted distance value of each block can also be predicted by applying median edge detector on distance values. For each block, after subtracting the distance value from the predicted distance value, the result value is embedded to the high and low reconstruction levels by using difference expansion. The experimental results show that the capacity of the proposed method is high and the quality of the image is good.