In　this　article,　we　present　the　optimization　of　a　neutron　imaging　system　for　Z-Pinch　driven　fusion　experiment　with　large　field　of　view　(FOV)　5　mm　in　diameter　and　low　neutron　yields　10(10)　to　10(12).　The　Inverse　Radius　Curvature　(IRC)　penumbra　aperture　design　method　was　employed　to　yield　thick　aperture　suitable　for　the　large　FOV.　A　preliminary　optimal　radius　of　the　aperture　which　makes　a　tradeoff　between　spatial　resolution　and　image　SNR　is　given.　The　dependence　of　imaging　signal　to　noise　(SNR)　and　isoplanatism　of　point　spread　function　(PSF)　on　the　aperture　thickness　is　discussed.　A　suitable　compromise　of　aperture　thickness　was　made　between　these　two　issues.　With　respect　to　the　scintillation　fiber　array,　the　spatial　resolution　of　the　array　is　influenced　by　the　diameter　of　fibers　and　the　range　of　recoil　protons.　The　relationships　between　fiber　length,　neutron　sensitivity,　temporal　response　and　crosstalk　effect　are　studied.　And　the　features　of　three　commonly　used　fiber　arrays　such　as　packing　fraction　and　trapping　fraction　are　taken　into　consideration　as　well.　The　optimized　imaging　system　was　simulated　using　GEANT4.　A　coded　image　of　＂E＂　character　with　2　x　10(10)　neutrons　is　acquired　and　unfolded　by　five　image　reconstruction　algorithms,　i.e.　Wiener　Filtering,　Richardson-Lucy,　Maximum　Entropy,　Total　Variation　and　Genetic　algorithm.　Only　Genetic　algorithm　shows　certain　robustness　in　this　condition.　The　reconstruction　performance　of　Genetic　algorithm　is　affected　by　the　neutron　yield.　(C)　2013　Elsevier　Ltd.　All　rights　reserved.