For　the　underwater　image　of　swimming　pool,　its　S/N　is　low　and　the　edges　are　fuzzy.　If　use　traditional　image　restoration　method　to　dispose　it　directly,　the　result　is　not　expected,　existing　weaknesses　such　as　enlarged　edge,　denoise　incompletely　and　so　on.　According　to　the　high　stabilization　of　robust　estimation　to　noise,　we　combined　robust　estimation　technique　with　image　restoration　method　together,　and　proposed　an　adaptive　robust　image　smoothing　algorithm.　This　method　set　up　an　image　model　based　on　robust　M-estimator,　and　saw　the　image　restoration　as　a　problem　of　position　estimation.　Experiments　with　underwater　images　of　swimming　pool　proved　that　the　proposed　approach　can　preserve　the　structure　information　in　the　original　image　while　smoothing　it　simultaneously.　©　2007　IEEE.