Abstract: Aiming at the security and secrecy of transmitting secret images through the Internet, combined with the multi-layered syndrome-trellis codes, this paper proposed a vector-quantization-based image-hiding method. First of all, according to the reconfigurable principle, multi-layered most significant bits of the cover image were used to generate a set of candidate blocks. Then a block matching process was used to find the best-matching block for each block of secret image from the set and record the index values, and the K-means clustering was used to select some representative blocks from the not-well-matched blocks as the best-matching blocks for them to substitute the unused candidate blocks in the block matching process. Furthermore, the Huffman coding scheme was used to compress the index values of the best-matching blocks, representative blocks, etc. Finally, the encoded data were embedded in the multi-layered least significant bits of the cover image using the multi-layered syndrome-trellis codes. The experimental results show that, compared with four previous methods, the visual quality of the stego images obtained by the proposed method is better, and the visual quality of the extracted images is also better; the proposed method can resist the detection of three common structural attacks; the proposed method exhibits the following advantages: a high embedding capacity; a good visual quality of the stego image; a small error between the extracted image and the secret image.