Ngenxa yokunyuka okuthe gqolo kwinqanaba lophuhliso lwetekhnoloji yolwazi, minyaka le amaxwebhu e-elektroniki alula ngakumbi kwaye afunwa ngokusetyenziswa kwaye aqala ukongamela kumajelo eendaba emveli. Ngoko ke, kubaluleke kakhulu ukuhlawula ingqalelo ngexesha elifanelekileyo ekukhuseleni umxholo wolwazi kungekuphela nje kumajelo eendaba emveli, kodwa nakumaxwebhu e-elektroniki. Yonke inkampani enkulu eneemfihlo zorhwebo, zikarhulumente kunye nezinye ifuna ukunqanda ukuvuza kolwazi olunokwenzeka kunye nokuthotyelwa kolwazi olucalulweyo, kwaye ukuba ukuvuza kufunyenwe, thatha amanyathelo okumisa ukuvuza kunye nokuchonga umaphuli.
Kancinci malunga neendlela zokukhusela
Ukwenza le misebenzi, izinto ezithile zokukhusela zingeniswa. Izinto ezinjalo zinokuba ziibhakhowudi, iithegi ezibonakalayo, iithegi ze-elektroniki, kodwa eyona nto inomdla kakhulu ziithegi ezifihliweyo. Omnye wabameli ababalaseleyo ziiwatermarks; Akusiyo imfihlo ukuba abashicileli babeka ii-watermark zabo (amachaphaza atyheli kunye namanye amanqaku) xa beprinta, kodwa siya kuthathela ingqalelo ezinye izinto zakudala ezinokubekwa kwiscreen sekhompyuter kwindawo yokusebenzela yomqeshwa. Izinto ezinjalo ziveliswa yiphakheji ekhethekileyo yesofthiwe edonsa izinto eziphathekayo phezulu kwendawo yokusebenza yomsebenzisi, ukunciphisa ukubonakala kwezinto eziphathekayo ngokwazo kwaye ngaphandle kokuphazamisa umsebenzi womsebenzisi. Ezi teknoloji zineengcambu zamandulo malunga nophuhliso lwezenzululwazi kunye neendlela ezisetyenziselwa ukubonisa ulwazi olufihliweyo, kodwa zinqabile kwihlabathi lanamhlanje. Le ndlela ifumaneka ikakhulu kwicandelo lomkhosi kunye nasephepheni, ukuchongwa ngokukhawuleza kwabasebenzi abangenangqondo. Ezi teknoloji ziqala ukuqaliswa kwindawo yorhwebo. Iiwatermark ezibonakalayo ngoku zisetyenziswa ngokukhutheleyo ukukhusela i-copyright yeefayile ezahlukeneyo zemidiya, kodwa ezingabonakaliyo zinqabile. Kodwa zikwavuselela umdla omkhulu.
Izixhobo zoKhuseleko
Ingabonakali ebantwini I-Watermarks zenza izinto ezahlukeneyo ezinokuthi, ngokomgaqo, zingabonakali emehlweni abantu, kwaye zinokuthi zigqunywe kumfanekiso ngendlela yamachaphaza amancinci kakhulu. Siza kuqwalasela izinto ezibonakalayo, kuba ezo zingabonakaliyo ngeliso zinokuba zingaphandle kwendawo yombala oqhelekileyo wabalindi abaninzi. Ezi zixhobo zinexabiso elithile ngenxa yezinga eliphezulu lokungabonakali. Nangona kunjalo, akunakwenzeka ukwenza ii-CEHs zingabonakali ngokupheleleyo. Kwinkqubo yokuphunyezwa kwabo, uhlobo oluthile lokuphazamiseka komfanekiso wesikhongozeli lungeniswa emfanekisweni, kwaye uhlobo oluthile lwezinto ezibonakalayo luvela kuyo. Makhe siqwalasele iindidi ezi-2 zezinto:
- Ngebhayisekile
- Umbhodamo (owaziswe kukuguqulwa komfanekiso)
Izinto ze-Cyclic zimela ulandelelwano oluthile olulinganiselweyo lwezinto eziphindaphindayo eziphinda ziphindwe kaninzi kumfanekiso weskrini (umzobo 1).
Izinto ze-chaotic artifacts zingabangelwa iintlobo ezahlukeneyo zokuguqulwa komfanekiso ogqithisiweyo (umzobo 2), umzekelo, ukuqaliswa kwehologram.
Irayisi. 1 Izinto ezenziwe ngebhayisekile
Irayisi. 2 Iziphithiphithi Artifacts
Okokuqala, makhe sijonge iinketho zokuqaphela ii-artifacts ze-cyclic. Ezo zixhobo zinokuba:
- iiwatermark zombhalo eziphinda kwiscreen sonke
- ukulandelelana kokubini
- iseti yamanqaku esiphithiphithi kwiseli yegridi nganye
Zonke izinto zakudala ezidwelisiweyo zisetyenziswa ngokuthe ngqo ngaphezulu komxholo obonisiweyo; Le ndlela ibandakanya ukusebenza kunye nendibaniselwano yeendawo ezigqithisileyo zengingqi nganye yetshaneli ye-histogram. Ingxaki ixhomekeke ekukhangeleni i-extrema yendawo kumfanekiso onzima kakhulu oneenkcukacha ezininzi eziguqukayo ezibukhali; Ungazama ukusebenzisa izihluzi ezahlukeneyo, kodwa ziyakwazisa ezazo iziphene, ezinokuthi ekugqibeleni zikhokelele ekungakwazini ukubona i-watermark. Kukwakho nokhetho lokuqaphela ezi zixhobo kusetyenziswa izixhobo ezithile zokubona edge (umzekelo, iCanny edge detector). Ezi ndlela zinendawo yazo yezinto zakudala ezibukhali kakhulu kwinguquko; abachongi banokuqaqambisa iikhonto zemifanekiso kwaye emva koko bakhethe uluhlu lwemibala ngaphakathi kwekhonto ukuhlanganisa umfanekiso ukuze baqaqambise ngakumbi izinto zakudala ngokwazo, kodwa ezi ndlela zifuna ukulungiswa ngokufanelekileyo ukuqaqambisa i-artifact iicontours ezifunekayo, kunye nokulandela okuhamba kabini komfanekiso ngokwawo xa kuthelekiswa nemibala ekwikhonto ezikhethiweyo. Ezi algorithms zithathwa njengezingathembekanga kwaye zizame ukusebenzisa uzinzile kwaye zizimeleyo kuhlobo lwamacandelo ombala womfanekiso.
Irayisi. 3 I-Watermark emva kokuguqulwa
Ngokubhekiselele kwizinto zakudala eziphazamisayo ezikhankanywe ngaphambili, i-algorithms yokuyiqonda iya kwahluka kakhulu. Ukususela ekubunjweni kwezinto zakudala ze-chaotic kuthathwa ngokubeka i-watermark ethile emfanekisweni, oguqulwa ngolunye utshintsho (umzekelo, i-discrete Fourier transform). I-artifacts evela kwiinguqu ezinjalo zisasazwa kwiscreen sonke kwaye kunzima ukuchonga ipateni yazo. Ngokusekwe koku, i-watermark iya kubekwa kuwo wonke umfanekiso ngendlela "engaqhelekanga" izinto zakudala. Ukuqatshelwa kwe-watermark enjalo yehla ukuya ngqo kwinguqu yomfanekiso usebenzisa imisebenzi yokuguqula. Isiphumo sokuguqulwa sinikezelwa kumzobo (umzobo 3).
Kodwa inani leengxaki ezivelayo ezithintela ukuqatshelwa kwe-watermark phantsi kweemeko ezifanelekileyo. Ngokuxhomekeke kuhlobo loguqulo, kunokubakho ubunzima obahlukeneyo, umzekelo, ukungenzeki kokuqondwa koxwebhu olufunyenwe ngokufota kwi-engile enkulu xa kuthelekiswa nescreen, okanye ifoto nje ekumgangatho ophantsi, okanye iscreen esigcinwe kuyo. ifayile enoxinzelelo oluphezulu lwelahleko. Zonke ezi ngxaki zikhokelela kwingxaki yokuchonga i-watermark; kwimeko yomfanekiso we-angle, kuyafuneka ukuba kufakwe iinguqu ezintsonkothileyo okanye kusetyenziswe iinguqu ezihambelanayo kumfanekiso, kodwa akukho namnye okanye omnye oqinisekisa ukubuyiswa ngokupheleleyo kwe-watermark. Ukuba siqwalasela imeko yokubamba isikrini, kuvela iingxaki ezimbini: eyokuqala kukugqwesa xa kuboniswa kwiscreen ngokwaso, okwesibini kukugqwethwa xa ugcina umfanekiso kwiscreen ngokwawo. Eyokuqala kunzima kakhulu ukuyilawula ngenxa yokuba kukho iimatrices zeemonitha zomgangatho owahlukileyo, kwaye ngenxa yokungabikho kombala omnye okanye omnye, bafaka umbala ngokuxhomekeke kumbala wabo, ngaloo ndlela bazisa ukugqwetheka kwi-watermark ngokwayo. Okwesibini kunzima ngakumbi, ngenxa yokuba ungagcina umfanekiso weskrini kuyo nayiphi na ifomathi kwaye, ngokufanelekileyo, ulahlekelwe yinxalenye yoluhlu lombala, ke ngoko, sinokuphulukana ne-watermark ngokwayo.
Iingxaki zokuphunyezwa
Kwihlabathi lanamhlanje, zininzi ii-algorithms zokwazisa ii-watermark, kodwa akukho nanye iqinisekisa i-100% yokwenzeka ngakumbi kwe-watermark emva kokuphunyezwa kwayo. Ubunzima obuphambili kukumisela isethi yeemeko zokuzala ezinokuvela kwimeko nganye ethile. Njengoko bekutshiwo ngaphambili, kunzima ukwenza i-algorithm yokuqaphela eya kuthatha ingqalelo kuzo zonke izinto ezinokubakho zokugqwetha kunye nokuzama ukonakalisa i-watermark. Umzekelo, ukuba isihluzo seGaussian sisetyenziswa kumfanekiso wangoku, kwaye izinto zakudala kumfanekiso wokuqala bezincinci kwaye zichasene nemvelaphi yomfanekiso, kuya kuba nzima ukuzazi, okanye inxalenye yewatermark iyakulahleka. . Makhe siqwalasele imeko yesithombe, kunye neqondo eliphezulu lokunokwenzeka liya kuba ne-moire (umzobo 5) kunye "negridi" (umzobo 4). I-Moire yenzeka ngenxa ye-discreteness ye-matrix yesikrini kunye ne-discreteness ye-matrix yezixhobo zokurekhoda kule meko, imifanekiso emibini ye-mesh ibekwe phezulu komnye nomnye. Umnatha uya kuthi ugqume ngokuyinxenye i-watermark artifics kwaye ubangele ingxaki yokuqonda i-moire, nayo, kwezinye iindlela zokuzinzisa i-watermark yenza ukuba kungenzeki ukuyiqaphela, kuba igqitha inxenye yomfanekiso ophawulweyo.
Irayisi. 4 Igridi yomfanekiso
Irayisi. 5 Moire
Ukuze ukwandise umda wokuqaphela ii-watermark, kuyimfuneko ukusebenzisa i-algorithms esekwe kwi-self-learning neural networks kunye nenkqubo yokusebenza, eya kufunda ngokwabo ukuqaphela imifanekiso ye-watermark. Ngoku kukho inani elikhulu lezixhobo zenethiwekhi ye-neural kunye neenkonzo, umzekelo, ukusuka kuGoogle. Ukuba uyanqwena, unokufumana iseti yemifanekiso yereferensi kwaye ufundise inethiwekhi ye-neural ukuqaphela izinto zakudala eziyimfuneko. Le ndlela inawona mathuba athembisayo okuchonga neewatermark ezigqwethekileyo, kodwa ukuchongwa ngokukhawuleza kufuna amandla amakhulu ekhompyuter kunye nexesha elide loqeqesho lokuchongwa okuchanekileyo.
Yonke into echazwayo ibonakala ilula, kodwa okukhona untywila kule miba, kokukhona uqonda ukuba ukuqonda ii-watermark kufuneka uchithe ixesha elininzi ekuphumezeni nayiphi na i-algorithms, kunye nexesha elingakumbi ekuziseni ukuba kunokwenzeka okufunekayo. ukuqonda umfanekiso ngamnye.
umthombo: www.habr.com
