Abstract: The data volume of stereo image increases exponentially, so the transmission and storage of image information encounter a great challenge. The self-organizing map is a very efficient clustering algorithm, it is widely researched and applied in image processing, data visualization and other fields. In order to improve the coding efficiency of the left image, this paper presents a stereo image coding algorithm based on the 3D frequency sensitive self-organizing map. Left image is encoded by the vector quantization based on 3DFSSOM in this algorithm. These residuals are encoded by the JPEG, which are obtained by the vector quantization and the disparity estimation. Disparity vector is encoded by the Huffman code. In addition, the proposed algorithm reduces the computational complexity of distortion measure by introducing inequality criterion in 3DFSSOM algorithm. The experiments show that coding efficiency of the left image is improved greatly, the computational complexity of algorithm is also reduced significantly.