To　address　the　problem　of　high　error　rate　in　matching　human　face＇s　low-texture　regions,　a　stereo　matching　algorithm　based　on　epipolar　distance　transform　is　presented.　Image　intensity　values　are　converted　to　a　relative　location　along　the　epipolar　line,　so　pixels　in　the　low-texture　regions　become　distinguishable,　which　improve　stereo　matching　algorithm　in　matching　precision　of　low-texture　regions.　When　the　method　is　applied　to　face　high　texture,　the　robustness　and　accuracy　of　stereo　matching　will　decline.　To　solve　the　problem,　favorable　parts　of　epipolar　distance　transform　results　are　extracted　in　different　epipolar　length.　The　results　of　experiments　show　that　higher　matching　accuracy　and　lower　computational　complexity　can　be　achieved　in　stereo　matching　of　human　face　with　this　transformation,　and　the　robustness　of　the　transform　for　high-texture　regions　is　demonstrated.　©,　2015,　Science　Press.　All　right　reserved.