Image coding plays a very important role in a digital TV system. In this paper we focus on the discussion of the image compression in DTV systems. We propose a regular, efficient algorithm, which has effectively much lower complexity. Our algorithm would require only 12 bits of coefficient wordlengths to achieve the CCITT/ISO specification. But the traditional method would need 12~18 bits to achieve the same specification. The hardware cost is about 35% higher than the traditional method. But the input block length to our design is 64, which is much higher than traditional algorithm, whose block length is generally between 8 and 16. This dictates that the throughput of our algorithm is about 2 to 4 time higher than the traditional algorithms.