Data　denoising　is　an　essential　issue　in　machine　learning　and　computer　vision.　However,　most　existing　denoising　methods　can　handle　only　low-noise　data,　because　it　is　difficult　for　these　methods　to　obtain　the　true　structures　of　high-noise　data.　An　existing　partial　solution　is　to　consolidate　the　structures　of　data　by　moving　each　sample　to　its　near　high　density　region.　It　actually　regards　all　non-noisy　samples　as　noisy　ones,　but　non-noisy　samples　are　important　for　constructing　the　structures　of　high-noise　data　so　they　should　not　be　moved.　To　address　this　problem,　we　propose　a　new　denoising　method　called　Denoising　by　a　new　Noise　index　and　Reinforcement　learning　(DNR).　Firstly,　it　detects　each　noisy　sample　by　its　density　and　the　distance　between　this　sample　and　the　center　of　its　neighbors.　A　noisy　sample　usually　has　a　low　density　and　most　neighbors　of　this　sample　will　be　on　the　same　side　of　it,　leading　to　the　center　of　its　neighbors　far　from　it,　especially　for　high-noise　data.　Secondly,　for　each　detected　noisy　sample,　DNR　models　its　movement　as　a　Markov　decision　process　to　store　the　experience　in　this　movement.　Finally,　NDR　learns　a　policy　to　iteratively　move　each　detected　noisy　sample　to　its　near　high　density　region　by　learns　the　experience　of　this　movement　in　reinforcement　learning.　The　learned　experience　can　effectively　help　the　movement　adapt　to　the　high-noise　in　real-world　cases.　In　DNR,　the　structures　of　high-noise　data　can　be　well　consolidated　by　our　detection　and　movement　of　noisy　samples.　To　prove　the　rationality　of　DNR,　we　theoretically　analyze　its　convergence.　Then,　we　perform　the　experiment　to　illustrate　that　DNR　can　better　denoise　high-noise　data　than　existing　denoising　methods.　The　source　code　can　be　downloaded　from　https://github.com/　TianyiHuang2022.　(c)　2022　Elsevier　Inc.　All　rights　reserved.