the　key　to　image　data　compression　is　extracting　main　feature　information　such　as　edge　and　mutation　part　of　the　image　signal.　In　order　to　improve　the　efficiency　of　image　data　compression　based　on　lifting　wavelet,　the　two　lifting　stage　such　as　prediction　and　update　can　realize　the　information　separation　from　high　frequency　to　low　frequency.　image　decomposition　is　completed　through　biorthogonal　wavelet　transform,　the　wavelet　coefficients　is　extracted　with　multi-scale　in　different　frequency　bands,　The　location,　dimension　and　corresponding　relationship　to　mutations　point　for　module　maximum　of　wavelet　coefficients　are　all　determined.　The　compression　process　is　stop　until　the　image　signal　can　be　approximately　reconstructed　from　these　feature　information,　image　feature　extraction　and　data　compression　are　realized　finally.　The　simulation　shows　that　the　lifting　wavelet　is　fully　competent　for　image　data　compression.　©　2010　IEEE.