As　the　steganography　technology　becomes　diversified　recently,　a　good　blind　steganalyzer　is　in　great　request.　In　recent　works,　a　blind　steganalysis　based　on　statistical　moments　of　wavelet　characteristic　functions　(MWCF)　is　proposed,　but　it　has　poor　generalization　ability　to　some　extent.　To　improve　this　weakness,　the　F-score　feature　selection　method　is　used　to　filter　irrelevant,　redundant　features　calculated　from　MWCF.　By　combining　MWCF　and　F-score　method,　an　improved　blind　steganalysis　method　is　proposed　in　this　article,　called　FS-MWCF　method　in　short.　Experimental　results　show　that　the　FS-MWCF　method　has　better　generalization　ability　and　lower　classifying　time　complexity,　less　than　half　of　that　using　MWCF　method.　©　2009　IEEE.