El Igama lesinqumo esikhulu selivele linendawo endaweni yalo mklamo ngenxa ye-Adobe Lapho yethulwa, lobu buchwepheshe obuthuthukisiwe obusekelwe ezigidini zezithombe abenzi lutho futhi obuvumela isithombe kusuka ku-10MP siye kwesinye se-40MP ngaphandle kokulahlekelwa imininingwane noma ubukhali.
Sengathi kungumlingo, lokhu Ubusha buka-Adobe bukhona buvela ku-Adobe Camera Raw e-Photoshop futhi kungekudala izofika kwezinye izinhlelo zokusebenza ezifana ne-Lightroom ne-Lightroom Classic. Sizochaza kancane ngaphezulu ukuthi iyini i-Adobe Super Resolution.
Yini i-Adobe Super Resolution
U-Eric Chan uphethe iphrojekthi yeSuper Resolution eyenziwayo nokuthi sesivele sinayo ku-Photoshop nge-Camera Raw. Sisho umphathi wephrojekthi wale ndaba entsha obesebenza ngaphambilini e-Photoshop enza izinto ezintsha ezinjenge: Shadows and Highlights, Clarity, Camera Profiles or Lens Correction phakathi kwamanye amaphrojekthi. Umsebenzi wakhe e-Photoshop uyabonakala futhi kulokhu uyasimangaza ngale nto entsha.
Ukulungiswa okuhle nakho uqobo kuyiphrojekthi ye-pixel, njengoba esho, kodwa kolunye uhlobo. Uyicacisa kahle ngokubeka ifayela le- isibonelo sesithombe esidala esinaso i-10MP kuthathwe ikhamera yesinqumo esiphansi sokuphrinta okukhulu. Futhi manje ake ucabange ukuthi unomsebenzi othuthukile "wokusondeza kwedijithali" ukukhulisa isihloko sesithombe esinjalo.
Futhi kungenxa yokuthi lokhu okusha kuvela isici esakhishwa eminyakeni eyi-2 eyedlule: Imininingwane ethuthukisiwe, nokuthi isebenzisa i-Machine Learning ukufaka amafayela aluhlaza ngezinga elithile lokwethembeka, okuholela emininingwaneni ebukhali enezinto ezimbalwa zobuciko.
Leyo Isixazululo esikhulu yisigaba sesibili salowo phrojekthi futhi lokho kubhekisa kunqubo yokwenza ngcono ikhwalithi yesithombe ngokuthuthukisa noma ngokwenza ngcono isinqumo sayo esibonakalayo.
Njengamanje nonke niyazi ukuthi ukukhulisa isithombe kubangela imininingwane efiphele ukuthi yenziwe, kepha iSuper Resolution iyakwazi ukuzinciphisa nge imodeli yokufunda yokufunda esezingeni eliphakeme noma iMishini yokufunda eqeqeshwe yizigidi zezithombe. Ngakho-ke iSuper Resolution iyakwazi ukukhulisa izithombe ngokukhaliphile ngenkathi igcina imiphetho ihlanzekile futhi igcina imininingwane ebaluleke kakhulu.
Manje umbuzo: ngabe sidinga amaphikseli amaningi?
Kuyacaca ukuthi ngaso sonke isikhathi uma sinamakhamera angcono nokuthi yini Ezinsukwini zakudala umuntu wayekwazi ukukhiqiza izithombe ze-6MP kwakubonakala kungaphezu kokwanele. Kepha-ke lawa ayi-12MP aba ngama- "6MP" amadala, futhi manje angama-24 angama-12. edlule futhi lokho akusakhulunywa ngalawo makhamera asuka ku-40 aya ku-100MP esithombeni esisodwa.
Impendulo, njengoba kusho uChan, ithi uma sidinga amaphikseli amaningi, ilula: ngezikhathi ezithile yebo. Uma sikusa emkhakheni wezithombe kusuka kumadivayisi eselula, sesivele sithola ingxenye ekwazi ukuthatha izithombe ze-12MP. Ngokunengqondo, kwezinye izithombe, ukusebenzisa i-Adobe Super Resolution kungaba cishe ebaluleke kakhulu. Ukukhulisa isithombe ngokusondeza kwedijithali ukusinika ukuthinta okucijile kungesinye sezibonelo ezinjalo.
Like buyisa izithombe eminyakeni eyedlule esazithatha ngekhamera esezingeni eliphansi, isibuyisela emuva ekutheni lesi sici esisha esethulwe yiChan nge-Adobe singaba lusizo kangakanani.
UChan ugcina ngokukhombisa ukuthi lesi sithombe esisinamathisele ukuthi leso sinqampuna sihlala kanjani ku-2.5MP futhi ufuna ukusiphrinta, ngeSuper Resolution angasiguqula sibe yi-10MP ukusithatha asiprinte asishiye silungile. Kunezibonelo eziningi esicabanga ngazo ngalo msebenzi weSuper Resolution ongazisebenzisa vele uma une-Cloud Cloud futhi uvuselele i-Adobe Photoshop nge-Camera Raw.
Isebenza kanjani iSuper Resolution
UMichaël Gharbi noRichard Zhang bakwa-Adobe Research babhekele intuthuko yengqikithi yeSuper Resolution. Umqondo uqhamuka ekuqeqesheni ikhompyutha enezigidi zamabhangqa ezithombe ezinesinqumo esiphansi nesiphezulu, ukuze ikhompyutha iqonde ukuthi zingakhuliswa kahle kanjani izithombe ezinesinqumo esiphansi ngemininingwane.
Lesi yisibonelo salokho:
Fundisa ikhompyutha ukuthi ifunde ezithombeni eziningi, njengoba kusho uChan, kufana nokufundisa ingane: umfundisa isakhiwo sokubulawa futhi unikele ngezigidi zezithombe njengezibonelo. Ku-Super Resolution lesi sakhiwo esiyisisekelo sithi ibizwa nge- "Deep Convolutional Neural Network", futhi leyo yindlela epholile yokusho ukuthi okwenzeka kuphikseli kuncike kulokho okuseduze kwayo. Lokhu kuqeqeshwa kungathatha izinsuku noma ngisho amasonto.
Ungayisebenzisa kanjani iSuper Resolution
Masifike ephuzwini:
- Qhafaza kwesokudla esithombeni noma ubambe isikhathi eside Ukhiye wokulawula ngenkathi uchofoza
- Khetha ukuthuthukisa ... kusuka kumenyu yokuqukethwe
- Ku Thuthukisa iwindi lokubuka kuqala, hlola ibhokisi leSuper Resolution
- Manje chofoza ekuthuthukiseni
- Linda kancane ngenkathi wenza izibalo futhi kuzokhiqizwa ifayili elisha eliluhlaza ku-Digital Negative equkethe isithombe esithuthukisiwe
- Manje usungahlela leso sithombe njenganoma yisiphi esinye
Kuyiqiniso lokho Ukulungiswa okuhle kunemikhawulo yayo njengoba izithombe zikhawulelwe kumaphikseli angama-65.000 kanye ne-500 MP. Uma uzama ukusebenzisa isithombe ngalezo zinombolo kuzonikeza iphutha.
I-Adjuntamos una uchungechunge lwezibonelo zokuthi kuthuthukiswa kanjani futhi okumangazayo:
Kuyinto a format nesithuthukisi esisebenza kuma-JPEGs, ama-PNG, nama-TIFFs, ngakho-ke unayo yonke inkululeko yokuthuthukisa zonke izinhlobo zezithombe ngale nsiza enkulu ku-Photoshop.