The　watermarking　capacity　of　a　digital　image　is　an　evaluation　of　how　much　information　can　be　hidden　within　digital　images.　This　paper　presents　a　blind　watermarking　algorithm　based　on　the　Hopfield　neural　network　and　discusses　the　bounds　of　watermark　information.　The　Hopfield　neural　network　is　used　to　store　the　original　image　during　the　watermark　embedding　and　to　retrieve　it　during　the　watermark　extracting.　According　to　the　research,　the　attraction　basin　of　neural　network　determines　the　watermarking　capacity.　©　Springer-Verlag　Berlin　Heidelberg　2005.