融合图像分割与风格迁移的文物数字拓片技术探索

  • 打印
  • 收藏
收藏成功


打开文本图片集

中图分类号:K87.4;TP391.41 文献标识码:A

Abstract:Inthefaceoftheburgeoningdemandforthedigitalizationofculturalheritage,traditionalmanual rubing technique has become increasingly inadequate in meting the high standard requirement of contemporaryculturalheritageprotectionandexhibition.Thisisatributabletoitsinherentrisks,intricate operationalprocedures,andchalengesinpreservation.Thenoveldigitalrubingtechnology,characterizedbyitseficiencyandsafety,hasmanifestedremarkableaplicationvaluewithintherealmofcultural heritageprotection,therebychartinganewcourseforthedigitalrepresentationofculturalheritageinformation.Toenhancethequalityandeficiencyofdigitalrubinggeneration,thispaperproposesaninovativeaproachgroundedonimagesegmentationandstyletransfer.Thismethodachievedahighdegre of integration betwen cultural heritage content and rubing style by integrating the precise image segmentationcapabilitiesoftheSegmentAnythingModel(SAM)withadeplearningbasedstyletransfernetwork fordigitalculturalheritagerubings.Specificaly,theSAM model,leveragingitsrobustgeneralization ability and Prompt-based interactive segmentation prowes, is capable of swiftly and acurately segmenting thetargetobjectswithinculturalheritageimages.Inthedomainofstyletransferfordigitalculturalheritagerubings,aconvolutionalneuralnetwork(CNN)architecturefoundedondeplearningisemployed. Throughtrainingtheneuralnetworktolearnthedistinctivefeaturesofspecificrubingstylesandaplying thesefeaturestotargetimages,digitalrubingimageswithanalogousstylesaregenerated.Thetechnical architectureincorporatesanencoder-decoderdesignandarubingstyleatentionmodule,andutilizesa detailpreservationlos functiontooptimizemodeltraining.Thisensuresthatthegenerateddigitalrubbingsnotonlypreservetheintricatedetailsofculturalrelicsbutalsofaithfulyreproducethestylisticcharacteristicsoftraditionalrubings.Experimentaloutcomesdemonstratethatthistechnologysignificantly elevatedthelevelofautomationandiscapableofrapidlyproducinghighqualitydigitalrubings.When comparedwithtraditionalimageprocesingand3Dreconstructiontechniques,thisnon-contactmethodobviatestheriskof hsicaldamaetoculturalrelics.Moreover,itstrikesafavorablebalancebetwendetail retentionandstyleconsistency,exhibitspotentialforlargescaleimplementation,andreducesoveral costs.Thistechnolo thusofersaneficientandinovativesolutionforthediitalrotectionanddiseminationofculturalheritage.

Keywords:digitalrubingtechnology;culturalheritagepreservation;imagesegmentation;styletransfer

拓片堪称中国传统捶拓技艺的精髓,它能够把碑刻、墓碣、摩崖石刻以及器物上的文字与图案非常精准地复制到宣纸之上,进而变成可以反映不同时代社会文化、政治、经济、宗教以及艺术等诸多方面信息的一种重要的文献载体[1]。(剩余12770字)

monitor
客服机器人