Recently,　Sparse　Representation　(SR)　models　have　made　a　great　success　in　the　field　of　face　recognition.　However,　when　face　image　has　occlusions　or　shows　varying　facial　expressions,　or　the　face　is　seriously　corrupted,　the　recognition　performance　of　the　existing　SR-based　methods　is　seriously　deteriorated.　To　address　this　problem,　in　this　paper　we　propose　a　new　novel　framework　RISR　(Robust　Iterative　Sparse　Reconstruction)　for　face　recognition.　In　our　proposed　framework,　the　disturbance　Effects　of　occlusions,　facial　expressions　and　corruptions,　have　been　Effectively　restricted　with　an　iterative　sparse　reconstruction　process.　We　conduct　extensive　comparison　experiments　on　simulated　corruption　face　dataset　and　real　varying　facial　expression　face　dataset,　and　the　results　illustrate　the　Effectiveness　our　proposed　RISR　model.　©,　2015,　Binary　Information　Press.　All　right　reserved.