{"id":955,"date":"2023-03-14T12:57:18","date_gmt":"2023-03-14T09:57:18","guid":{"rendered":"https:\/\/legendhomework.com\/tr\/?p=955"},"modified":"2023-03-14T12:57:18","modified_gmt":"2023-03-14T09:57:18","slug":"goruntulerin-cozunurlugu-multimedya-bolumu-multimedya-bolumu-odevleri-multimedya-bolumu-tez-yaptirma-multimedya-bolumu-odev-ucretleri","status":"publish","type":"post","link":"https:\/\/legendhomework.com\/tr\/goruntulerin-cozunurlugu-multimedya-bolumu-multimedya-bolumu-odevleri-multimedya-bolumu-tez-yaptirma-multimedya-bolumu-odev-ucretleri\/","title":{"rendered":"G\u00f6r\u00fcnt\u00fclerin \u00c7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fc \u2013 Multimedya B\u00f6l\u00fcm\u00fc \u2013 Multimedya B\u00f6l\u00fcm\u00fc \u00d6devleri \u2013 Multimedya B\u00f6l\u00fcm\u00fc Tez Yapt\u0131rma \u2013Multimedya B\u00f6l\u00fcm\u00fc \u00d6dev \u00dccretleri"},"content":{"rendered":"

\u00d6\u011frenilmi\u015f \u00c7oklu Rekonstr\u00fcksiyon \u00c7ekirde\u011fi ile G\u00f6r\u00fcnt\u00fclerin \u00c7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fc<\/span><\/strong><\/h4>\n

S\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck, fiziksel sens\u00f6r taraf\u0131ndan sa\u011flanandan daha y\u00fcksek bir \u00e7\u00f6z\u00fcn\u00fcrl\u00fc\u011fe ula\u015fan sinyal i\u015fleme i\u015flemine verilen terimdir. Bu terim, radar toplulu\u011fu i\u00e7inde yayg\u0131nd\u0131r ve uzayda ayr\u0131lm\u0131\u015f nesneleri radar taraf\u0131ndan sa\u011flanan \u00e7\u00f6z\u00fcn\u00fcrl\u00fckten daha az ay\u0131rt etme yetene\u011fini i\u00e7erir. Optik imgeler alan\u0131ndaki sorun, m\u00fckemmel yeniden yap\u0131land\u0131rma sorununa benzer.<\/span><\/p>\n

Bu nedenle, bu b\u00f6l\u00fcm sonlu bir \u00f6rnek k\u00fcmesinden g\u00f6r\u00fcnt\u00fc b\u00fcy\u00fctme konusunu (interpolasyon, yak\u0131nla\u015ft\u0131rma, b\u00fcy\u00fctme, vb. olarak da an\u0131l\u0131r) ele alacakt\u0131r. B\u00fcy\u00fctmenin multimedya uygulamalar\u0131na yard\u0131mc\u0131 olabilece\u011fi bir \u00f6rnek, bit h\u0131z\u0131 k\u0131s\u0131tlamalar\u0131n\u0131n video \u00e7\u0131k\u0131\u015f\u0131n\u0131 s\u0131n\u0131rlad\u0131\u011f\u0131 video telekonferanst\u0131r.<\/span><\/p>\n

Bu t\u00fcr k\u0131s\u0131tlamalar, tipik olarak, d\u00fc\u015f\u00fck g\u00f6rsel kaliteye sahip y\u00fcksek oranda s\u0131k\u0131\u015ft\u0131r\u0131lm\u0131\u015f ve boyutla s\u0131n\u0131rl\u0131 bir video dizisinin iletilmesine neden olur. Bu ba\u011flamda, s\u00fcper-\u00e7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fcn g\u00f6revi b\u00f6ylece, g\u00f6r\u00fcnt\u00fclerin s\u0131k\u0131\u015ft\u0131r\u0131lm\u0131\u015f dizisine kay\u0131p bilgiyi geri y\u00fcklemek ve b\u00f6ylece bunlar\u0131n b\u00fcy\u00fct\u00fclmesinin yan\u0131 s\u0131ra daha keskin ve\/veya daha az bozulmu\u015f bir g\u00f6r\u00fcnt\u00fc sa\u011flamak olur.<\/span><\/p>\n

Multimedyan\u0131n s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fckten yararlanabilece\u011fi alanlar aras\u0131nda, burada odaklan\u0131lan nokta hareketsiz g\u00f6r\u00fcnt\u00fclerin s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fc ile sonu\u00e7lanan g\u00f6r\u00fcnt\u00fc i\u015flemedir. \u00d6nerilen mimarinin sa\u011flad\u0131\u011f\u0131 faydalar incelenecek ve metodoloji ile ilgili \u00e7e\u015fitli konular tart\u0131\u015f\u0131lacakt\u0131r.<\/span><\/p>\n

Genel olarak, g\u00f6r\u00fcnt\u00fc b\u00fcy\u00fctme, g\u00f6r\u00fcnt\u00fc \u00f6rneklerinin \u00e7ift do\u011frusal, \u00e7ift k\u00fcbik veya k\u00fcbik B-spline \u00e7ekirdekleri gibi tek bir \u00e7ekirdekle konvol\u00fcsyonu yoluyla ger\u00e7ekle\u015ftirilir ve herhangi bir son i\u015fleme veya m\u00fcteakip g\u00f6r\u00fcnt\u00fc geli\u015ftirme tipik olarak ad hoc bir \u015fekilde ger\u00e7ekle\u015ftirilir.<\/span><\/p>\n

\u00d6rt\u00fc\u015fme veya di\u011fer olgulardan kaynaklanan kusurlar\u0131n bu tip do\u011frusal filtrelemeyle azalt\u0131lmas\u0131 \u00e7ok s\u0131n\u0131rl\u0131d\u0131r. Daha yak\u0131n zamanlarda, g\u00f6r\u00fcnt\u00fc alan\u0131 bilgisine dayal\u0131 b\u00fcy\u00fctme teknikleri ara\u015ft\u0131rma konusu olmu\u015ftur. \u00d6rne\u011fin, y\u00f6nl\u00fc y\u00f6ntemler bir g\u00f6r\u00fcnt\u00fcn\u00fcn yerel kenar i\u00e7eri\u011fini inceler ve y\u00fcksek frekans y\u00f6n\u00fc (kenar boyunca) yerine d\u00fc\u015f\u00fck frekans y\u00f6n\u00fcnde (kenar boyunca) enterpolasyon yapar.<\/span><\/p>\n

Birden \u00e7ok \u00e7ekirdek y\u00f6ntemi tipik olarak birka\u00e7 ad hoc enterpolasyon \u00e7ekirde\u011fi aras\u0131nda se\u00e7im yapar. Ortogonal d\u00f6n\u00fc\u015f\u00fcm y\u00f6ntemleri, ayr\u0131k kosin\u00fcs d\u00f6n\u00fc\u015f\u00fcm\u00fcn\u00fcn (DCT) ve dalgac\u0131k d\u00f6n\u00fc\u015f\u00fcm\u00fcn\u00fcn kullan\u0131m\u0131na odaklan\u0131r. Varyasyonel y\u00f6ntemler, enterpolasyon problemini bir fonksiyonelin k\u0131s\u0131tl\u0131 minimizasyonu olarak form\u00fcle eder. Bu y\u00f6ntemleri uzun uzad\u0131ya tart\u0131\u015fan geni\u015fletilmi\u015f bir literat\u00fcr taramas\u0131 sa\u011flanm\u0131\u015ft\u0131r.<\/span><\/p>\n

Burada sunulan yakla\u015f\u0131m yenidir ve benzer (ili\u015fkili) kom\u015fuluklar\u0131n \u00f6l\u00e7ekler aras\u0131nda benzer kald\u0131\u011f\u0131n\u0131 ve bu a priori yap\u0131n\u0131n \u00f6l\u00e7ekler aras\u0131nda mevcut g\u00f6r\u00fcnt\u00fc \u00f6rneklerinden yerel olarak \u00f6\u011frenilebilece\u011fini varsayarak s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fcn hatal\u0131 do\u011fas\u0131n\u0131 ele al\u0131r.<\/span><\/p>\n

Bu t\u00fcr yerel bilgi \u00e7\u0131karma, tipik olarak bir g\u00f6r\u00fcnt\u00fcn\u00fcn \u00f6rt\u00fc\u015fmeyen alt bloklar\u0131 k\u00fcmesini s\u0131k\u0131\u015ft\u0131ran JPEG ve PCA tabanl\u0131 yakla\u015f\u0131mlar\u0131n g\u00f6sterdi\u011fi gibi, g\u00f6r\u00fcnt\u00fc s\u0131k\u0131\u015ft\u0131rma \u015femalar\u0131nda olduk\u00e7a uzun bir s\u00fcredir \u00f6ne \u00e7\u0131k\u0131yor.<\/span><\/p>\n

\n

Ekran \u00e7\u00f6z\u00fcn\u00fcrl\u00fckleri<\/a><\/span>
\nPhotoshop \u00e7\u00f6z\u00fcn\u00fcrl\u00fck ka\u00e7 olmal\u0131<\/span>
\n2K \u00e7\u00f6z\u00fcn\u00fcrl\u00fck<\/span>
\nUHD \u00e7\u00f6z\u00fcn\u00fcrl\u00fck<\/span>
\nPhotoshop \u00e7\u00f6z\u00fcn\u00fcrl\u00fck ayar\u0131<\/span>
\nQHD \u00e7\u00f6z\u00fcn\u00fcrl\u00fck<\/span>
\n1024×768 \u00e7\u00f6z\u00fcn\u00fcrl\u00fck nas\u0131l yap\u0131l\u0131r<\/span>
\n4K \u00e7\u00f6z\u00fcn\u00fcrl\u00fck<\/span><\/p>\n

\n

Son s\u0131k\u0131\u015ft\u0131rma yakla\u015f\u0131mlar\u0131, alt bloklar aras\u0131ndaki bloklar aras\u0131 korelasyondan da yararlan\u0131r. Ama\u00e7, alt bloklar k\u00fcmesini, orijinal k\u00fcmeden daha verimli bir \u015fekilde bireysel olarak temsil edilebilecek s\u0131n\u0131rl\u0131 say\u0131da ayr\u0131k k\u00fcmeye b\u00f6lmektir.<\/span><\/p>\n

Benzer \u00f6rt\u00fc\u015fen mahalleleri y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc muadillerine e\u015flemek i\u00e7in bloklar aras\u0131 korelasyondan yararland\u0131\u011f\u0131m\u0131z i\u00e7in yakla\u015f\u0131m\u0131m\u0131z ruhen benzerdir. Ancak, bizden \u00f6nce hi\u00e7 kimse bu bilgiyi g\u00f6r\u00fcnt\u00fclerin yerini alabilecek k\u0131s\u0131tlamalar olu\u015fturmak i\u00e7in kullanmay\u0131 \u00f6nermedi.<\/span><\/p>\n

Ayr\u0131ca, \u00e7ok basit bir yerel mimarinin bu yap\u0131y\u0131 etkili bir \u015fekilde \u00f6\u011frenebilece\u011fini g\u00f6steriyoruz. Dahas\u0131, yakla\u015f\u0131m\u0131m\u0131z\u0131n, geleneksel \u00f6rnekleme teorisi kavramlar\u0131n\u0131n bir uzant\u0131s\u0131n\u0131 temsil eden, mevcut g\u00f6r\u00fcnt\u00fclerden olu\u015fturulmu\u015f ve yerel \u00f6zelliklerine “ayarlanm\u0131\u015f” bir \u00e7ekirdek ailesi ile bir evri\u015fime e\u015fde\u011fer oldu\u011fu g\u00f6sterilmi\u015ftir.<\/span><\/p>\n

B\u00f6l\u00fcm a\u015fa\u011f\u0131daki gibi b\u00f6l\u00fcmlere ayr\u0131lm\u0131\u015ft\u0131r. Burada tart\u0131\u015f\u0131lan s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fc kavramsal olarak tan\u0131t\u0131r. Yorumlar ve g\u00f6zlemler yap\u0131l\u0131r ve yerel mimarinin do\u011fdu\u011fu metodoloji de tarif edilir.<\/span><\/p>\n

D\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc g\u00f6r\u00fcnt\u00fclerimizi sentezlemek i\u00e7in kullan\u0131lan g\u00f6r\u00fcnt\u00fc elde etme modelini sunar. B\u00fcy\u00fctme i\u00e7in tek ve \u00e7ok \u00e7ekirdekli tabanl\u0131 yakla\u015f\u0131mlar\u0131 a\u00e7\u0131klar. S\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck metodolojisini uygulayan yerel mimariyi detayland\u0131r\u0131r. Mimarinin kapasitesini g\u00f6steren \u00e7e\u015fitli sonu\u00e7lar sunulmu\u015ftur. Metodoloji ile ilgili \u00e7e\u015fitli konular\u0131 tart\u0131\u015f\u0131r ve sonu\u00e7lar\u0131m\u0131z\u0131 sunar.<\/span><\/p>\n

S\u00fcper \u00c7\u00f6z\u00fcn\u00fcrl\u00fc\u011fe Bir Yakla\u015f\u0131m<\/span><\/strong><\/h3>\n

Burada sunulan s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck yakla\u015f\u0131m\u0131, Shannon \u00f6rnekleme teorisi taraf\u0131ndan dayat\u0131lan s\u0131n\u0131r\u0131n \u00f6tesinde bir g\u00f6r\u00fcnt\u00fcn\u00fcn (sonlu bir \u00f6rnek k\u00fcmesinden) yeniden olu\u015fturulmas\u0131n\u0131 ele al\u0131r. Basit olmas\u0131 ad\u0131na, geli\u015ftirmemiz tek boyutlu (1D) sinyaller kullan\u0131r, ancak iki boyutlu (2D) uzant\u0131lar a\u00e7\u0131k olmal\u0131d\u0131r.<\/span><\/p>\n

x(t), burada \u2212\u221e < t < \u221e, maksimum frekans i\u00e7eri\u011fi c rad\/s olan s\u00fcrekli bir sinyal olsun. Bu nedenle, analizimiz bant s\u0131n\u0131rl\u0131 sinyallere dayanmaktad\u0131r. x(t)’yi bir dizi temel fonksiyonun lineer kombinasyonu olarak \u015fu \u015fekilde g\u00f6sterebiliriz:<\/span><\/p>\n

Do\u011frusal a\u011f\u0131rl\u0131kland\u0131rman\u0131n x[n]’deki \u00f6rnekler taraf\u0131ndan verildi\u011fi ve k(t,n)’nin temel fonksiyonlar dizimizi temsil etti\u011fi yer. Burada x[n] \u2261 x(nTs), \u2212\u221e < n < \u221e’yi sa\u011flayan n tamsay\u0131lar\u0131 ve Ts \u00f6rnekleme periyodu i\u00e7in. \u00d6rnekleme teorisinde bir sinyalin m\u00fckemmel \u015fekilde yeniden yap\u0131land\u0131r\u0131lmas\u0131n\u0131 a\u00e7\u0131klayan denklem, tan\u0131m\u0131 gere\u011fi sin c(t) \u2261 sin(\u03c0t) oldu\u011fu yerdir. Temel fonksiyonlar\u0131m\u0131z\u0131n k(t, n) = t \u03c0t sin c( Ts \u2212 n) taraf\u0131ndan verildi\u011fini ve bu sonsuz k\u00fcmedeki temel fonksiyonlar\u0131n ortogonal oldu\u011funu g\u00f6r\u00fcyoruz.<\/span><\/p>\n

Bu nedenle, x(t)’nin her \u00f6rne\u011fi, \u00f6rneklerin yo\u011funlu\u011funun yeterince y\u00fcksek olmas\u0131 ko\u015fuluyla sonsuz say\u0131da \u00f6rnekle tam olarak \u00e7\u00f6z\u00fclebilir. \u00d6rnekleme periyodu Ts, sinyallerimizin ortogonal bir do\u011frusal izd\u00fc\u015f\u00fcm k\u00fcmesinden ve sonsuz say\u0131da \u00f6rnekten m\u00fckemmel bir \u015fekilde yeniden olu\u015fturulabilece\u011fi (\u00e7\u00f6z\u00fcmlenebilece\u011fi) s\u0131n\u0131r\u0131 sa\u011flar.<\/span><\/p>\n

Sin c bazlar\u0131n\u0131n, zaman (uzay) i\u00e7inde bant-s\u0131n\u0131rl\u0131 sinyalleri m\u00fckemmel bir \u015fekilde yeniden olu\u015fturabilen evrensel do\u011frusal izd\u00fc\u015f\u00fcmler seti oldu\u011funa dikkat edin. Bu temeller dizisi, i\u00e7eriklerine bak\u0131lmaks\u0131z\u0131n, uygun \u015fekilde \u00f6rneklenmi\u015f t\u00fcm sonsuz \u00f6l\u00e7\u00fcde bant s\u0131n\u0131rl\u0131 sinyallerin bunlarla yeniden olu\u015fturulabilmesi a\u00e7\u0131s\u0131ndan evrenseldir.<\/span><\/p>\n

Sonlu kapsam verileri durumu, \u00f6zellikle g\u00f6r\u00fcnt\u00fcler i\u00e7in daha ger\u00e7ek\u00e7idir. Sonlu \u00f6l\u00e7\u00fcdeki veriler i\u00e7in denklem ifade edilebilir.<\/span><\/p>\n

N, penceremizin kapsam\u0131d\u0131r. x\u02c6’deki \u015fapka \u00fcst simgesinin, sonlu say\u0131da veri \u00f6rne\u011finin sa\u011flad\u0131\u011f\u0131 yakla\u015f\u0131m\u0131 belirtmek i\u00e7in kas\u0131tl\u0131 olarak kullan\u0131ld\u0131\u011f\u0131na dikkat edin. Sonlu veri seti, sinyalin \u00e7\u00f6z\u00fclebilirli\u011fini azalt\u0131r. Bunu (4.3) denklemini frekans alan\u0131nda inceleyerek g\u00f6rebiliriz. Denklemin (4.3) s\u00fcrekli Fourier d\u00f6n\u00fc\u015f\u00fcm\u00fc verir.<\/span><\/p>\n

\n\n
<\/div>\n\n<\/div>\n \n