Image compression becomes the main task for reducing the image bit rate with acceptable reconstructed image quality. Wavelet transform shows a good way of subdividing the image into different subbands with different resolutions which are quantized to reduce the image bit rate. This paper introduces the application of Vector qauntization (VQ) to the image subbands in the wavelet domain. The image spectrum is decomposed into hierarchical subbands using the wavelet transform. Three methods of codebook generation are discussed namely, global codebook, local codebook, and variance based codebook with application to image subbands for different block sizes. The quantized coefficients and the baseband are Hufmann coded. Results showed that the local codebook based method gave better peak signal to noise ratio (PSNR) than the global codebook with nearly the same compression ratio (CR). But for variance based method, the visual quality of the reconstructed image is better than those of the two other methods for nearly the same PSNR, and CR as the edges are preserved.