Text　images　captured　by　the　surveillance　system　or　hand-hold　cameras　often　suffer　from　motion　blur　due　to　the　complex　relative　motion　between　the　camera　and　the　target　during　the　exposure　time.　The　accuracy　of　the　kernel　estimation　and　the　effective　priors　for　clear　text　images　are　two　important　keys　in　blind　motion　deblurring.　A　novel　blind　motion　deblurring　algorithm　is　proposed　for　text　images　in　a　complex　scene.　A　criterion　for　selecting　informative　edges　and　an　L-0　constraint　are　combined　to　improve　the　accuracy　of　the　kernel　estimation.　Then,　a　fast　non-blind　deconvolution　scheme　is　applied　to　accelerate　the　algorithm.　Experimental　results　on　text　images　show　that　the　proposed　method　can　achieve　high-quality　results　with　low　computational　complexity.