Discrete cosine transform (DCT) is an excellent transform because of its simple realization and high efficiency of compression. However, DCT cannot be combined with the fast and efficient zero-block coding algorithm perfectly. A new image coefficient distribution model using DCT is proposed based on the embedded zero-block coding algorithm. This model can make the same image coefficient distribution of DCT as that of discrete wavelet transform by coefficient clustering. During coding, the quadtree partitioning scheme of embedded zero-block coding is optimized. Experiment results show that the new algorithm can not only reduce the coding complexity, but also provides peak signal to noise ratio (SNR) performance better than baseline joint picture expert group (JPEG) at the same output bit rates.