The integer wavelet transform (IWT) is dealt with. For for image compression performance, it is known that the IWT has a worse performance than the discrete wavelet transform (DWT). We analyse the IWT signal representation mechanism with respect to the DWT one, in order to explain this difference. Insight into this mechanism suggests a simple, computationally efficient yet very effective way to improve the IWT compression performance up to the DWT level. Simulation results are presented for both natural and synthetic images, these latter being more critical for the IWT; it is shown that the proposed method permits one to fill the gap between the IWT and the DWT performance. © 1999 IEEE.