A hybrid image compression algorithm consisting of vector quantization followed with an entropy coder to code the truncated residual is described. It is demonstrated that the tree level decision rule for a hybrid adaptive vector quantization algorithm can be more efficient by including coding information from both the VQ and entropy coders. A cost function approach is taken where the vector cost is a sum of the total distortion for that vector and α times the total bit length from both the VQ and entropy encoders. An LMS method for adapting α is shown and the results summarized.