In　recent　years,　as　a　means　of　identity　authentication,　the　finger-vein　pattern　has　gradually　attracted　widespread　attention　in　the　academia　and　industry　due　to　its　unique　biological　characteristics,　high　stability,　and　wide　applicability.　Although　some　progress　has　been　made　in　finger-vein　recognition,　most　existing　methods　typically　rely　on　much　experience　and　require　numerous　labelled　training　samples.　In　order　to　break　through　this　limitation,　this　paper　develops　a　novel　Compact　Binary　Code　Learning　(CBCL)　method　for　few-shot　finger　vein　identification.　Specifically,　a　set　of　linear　projection　functions　are　jointly　learned　by　training　finger-vein　images　to　convert　texture　information　features　into　discriminant　binary　codes.　Subsequently,　the　learned　binary　codes　are　weighted　and　summed　to　real　values　to　generate　the　feature　map,　which　realizes　the　adaptive　learning　and　coding　of　the　finger　vein　features.　Following　this,　we　calculate　the　local　block　histograms　of　each　feature　map　and　integrate　them　into　a　feature　vector　as　the　final　feature　representation.　For　testing,　the　texture　features　of　the　finger　vein　testing　images　are　mapped　to　binary　features　for　matching　based　on　a　pre-trained　projection　matrix.　Experiments　results　on　two　SDUMLA　and　CAUC　finger-vein　databases　show　that　the　proposed　CBCL　method　is　superior　to　the　current　state-of-the-art　finger　vein　recognition　method,　demonstrating　its　powerful　feature　representation　ability　in　few-shot　learning.　©　The　Author(s),　under　exclusive　license　to　Springer　Nature　Singapore　Pte　Ltd.　2025.