Dual　energy　computed　tomography　(DECT)　has　significant　impacts　on　material　characterization,　bone　mineral　density　inspection,　nondestructive　evaluation　and　so　on.　In　spite　of　great　progress　has　been　made　recently　on　reconstruction　algorithms　for　DECT,　there　still　exist　two　main　problems:　1)　For　polyenergetic　X-ray　source,　the　tube　spectrum　needed　in　reconstruction　is　not　always　available.　2)　The　reconstructed　image　of　DECT　is　very　sensitive　to　noise　which　demands　special　noise　suppression　strategy　in　reconstruction　algorithm　design.　In　this　paper,　we　propose　a　novel　method　for　DECT　reconstruction　that　reconstructs　tube　spectrum　from　projection　data　and　suppresses　image　noise　by　introducing　l(1)-norm　based　regularization　into　statistical　reconstruction　for　polychromatic　DECT.　The　contribution　of　this　work　is　twofold.　1)　A　three　parameters　model　is　devised　to　represent　spectrum　of　ployenergetic　X-ray　source.　And　the　parameters　can　be　estimated　from　projection　data　by　solving　an　optimization　problem.　2)　With　the　estimated　tube　spectrum,　we　propose　a　computation　framework　of　l(1)-norm　regularization　based　statistical　iterative　reconstruction　for　polychromatic　DECT.　Simulation　experiments　with　two　phantoms　were　conducted　to　evaluate　the　proposed　method.　Experimental　results　demonstrate　the　accuracy　and　robustness　of　the　spectrum　model　in　terms　of　that　comparable　reconstruction　image　quality　can　be　achieved　with　the　estimated　and　ideal　spectrum,　and　validate　that　the　proposed　method　works　with　attractive　performance　in　terms　of　accuracy　of　reconstructed　image.　The　root　mean　square　error　(RMSE)　between　the　reconstructed　image　and　the　ground　truth　image　are　7.648　x　10(-4)　and　2.687　x　10(-4)　for　the　two　phantoms,　respectively.