According　to　the　characteristics　of　domestic　medium-resolution　remote　sensing　data,　we　proposed　a　new　method　for　retrieving　aerosol　optical　depth　(AOD)　from　optical　images　by　integrating　the　advantages　of　dark-pixel　and　deep-blue　algorithms.　The　method　was　proved　to　be　effective　and　suitable　for　AOD　estimation　over　both　low-reflectivity　and　high-reflectivity　surface　cover　types.　Here,　we　estimated　AOD　at　550　nm　over　Taiwan　Island　based　on　GF-1　WFV　data,　and　the　results　were　compared　with　MODIS　AOD　products　(MOD04).　The　spatial　distributions　of　AOD　from　GF-1　are　well　consistent　with　those　from　MOD04　(r＞0.9).　Moreover,　the　inversion　results　were　validated　by　using　the　ground-measured　AODs　from　AERONET.　The　retrieved　AOD　is　significantly　correlated　with　the　AERONET　data　(r＞0.85),　and　about　70%　of　the　estimated　results　meet　the　accuracy　requirement.　Compared　with　the　dark-pixel　and　deep-blue　algorithms,　the　proposed　method　has　distinctive　advantages　in　terms　of　estimation　accuracy　and　spatial　coverage.　In　Taiwan　Island,　the　marine　and　continental　aerosol　types　are　suitable　for　AOD　estimation　in　summer　and　winter,　respectively.　With　the　decrease　of　spatial　resolution　of　GF-1　images,　the　correlation　coefficient　between　the　image-estimation　and　ground-observation　AOD　presents　a　trend　of　decline　first　and　then　stable,　while　the　overall　change　is　not　significant.　©　2018,　Science　Press.　All　right　reserved.