The　code　length　of　Low-Density　Parity-Check　(LDPC)　Codes　is　usually　very　long,　and　a　lot　of　data　storage　and　computing　are　required　for　decoding,　so　the　method　and　the　level　of　quantization　used　in　messages　of　decoding　algorithm　play　a　key　influence　on　decoding　performance　and　hardware　resource　consumption.　Based　on　analyzing　various　key　data　in　the　Min-Sum　decoding　algorithm　and　the　particularity　of　the　algorithm,　a　non-uniform　all-integer　quantization　method　for　this　algorithm　is　proposed　in　this　paper.　Fewer　quantization　length　of　the　messages　required　and　all　integer　operations　were　used　in　decoding,　the　method　proposed　greatly　reduce　the　hardware　complexity,　and　improves　the　decoding　performance　by　providing　more　quantization　levels　for　low　level　messages.　Simulation　results　show　that　using　6-bit　non-uniform　quantized　messages　suffers　a　bit　error　rate　(BER)　degradation　of　only　about　0.05　dB　at　a　BER　of　10-7　compared　to　using　32-bit　floating-point　messages,　and　it　can　save　3-4　bit　compared　to　using　uniform　quantization　while　still　achieving　the　same　level　of　BER　performance.