Missing　data　have　emerged　in　broad　disciplines　such　as　biology,　geophysics,　economics,　public　health,　and　social　science.　This　article　explores　the　optimal　estimation　of　high-dimensional　covariance　matrix　with　missing　data　over　a　general　sparse　space　H　epsilon(cn,　p).　First,　the　upper　bounds　of　adaptive　entrywise　thresholding　estimator　are　proposed.　Then　the　minimax　lower　bound　is　established　by　a　simple　and　effective　approach.　Finally,　numerical　simulations　and　real　data　analysis　demonstrate　the　advantages　of　our　estimator　Sigma　tau　over　the　estimator　Sigma　at　of　Cai　and　Zhang　(2016).