{"id":961,"date":"2023-03-14T16:02:03","date_gmt":"2023-03-14T13:02:03","guid":{"rendered":"https:\/\/legendhomework.com\/tr\/?p=961"},"modified":"2023-03-14T16:02:03","modified_gmt":"2023-03-14T13:02:03","slug":"goruntu-edinme-modeli-multimedya-bolumu-multimedya-bolumu-odevleri-multimedya-bolumu-tez-yaptirma-multimedya-bolumu-odev-ucretleri","status":"publish","type":"post","link":"https:\/\/legendhomework.com\/tr\/goruntu-edinme-modeli-multimedya-bolumu-multimedya-bolumu-odevleri-multimedya-bolumu-tez-yaptirma-multimedya-bolumu-odev-ucretleri\/","title":{"rendered":"G\u00f6r\u00fcnt\u00fc Edinme Modeli \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":"

Metodolojinin Tan\u0131m\u0131<\/span><\/strong><\/h3>\n

Daha \u00f6nce ima edildi\u011fi gibi, s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck sorunu, “daha iyi” yeniden yap\u0131land\u0131r\u0131lm\u0131\u015f bir g\u00f6r\u00fcnt\u00fc elde etmek i\u00e7in toplanan bir dizi \u00f6rne\u011fe uygulanan uygun bir e\u015flemenin belirlenmesi sorunudur. Bu e\u015flemenin belirlenme bi\u00e7imi, ortaya \u00e7\u0131kan s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck s\u00fcrecini a\u00e7\u0131klar.<\/span><\/p>\n

Burada sunulan metodoloji, hatal\u0131 s\u00fcper\u00e7\u00f6z\u00fcn\u00fcrl\u00fck sorununa bir \u00e7\u00f6z\u00fcm elde etmek i\u00e7in gerekli ek bilgileri (yani toplanan numunelerin \u00f6tesinde) \u00e7\u0131karmak i\u00e7in yukar\u0131da belirtilen g\u00f6r\u00fcnt\u00fc \u00f6zelliklerinden yararlanarak s\u00fcper\u00e7\u00f6z\u00fcn\u00fcrl\u00fck ger\u00e7ekle\u015ftirir.<\/span><\/p>\n

P\u00fcr\u00fczs\u00fczl\u00fck varsaymak veya di\u011fer tipik k\u0131s\u0131tlamalara g\u00fcvenmek yerine, belirli bir g\u00f6r\u00fcnt\u00fc s\u0131n\u0131f\u0131n\u0131n yerel olarak benzer bilgiler i\u00e7erdi\u011fi ve bu benzerli\u011fin \u00f6l\u00e7ekler aras\u0131nda ge\u00e7erli oldu\u011fu ger\u00e7e\u011fini kullan\u0131r\u0131z. Dolay\u0131s\u0131yla temel sorun, yerel bilgilerin benzerli\u011fini belirleyebilecek ve \u00f6l\u00e7ekler aras\u0131ndaki benzerlikleri otomatik bir \u015fekilde yakalayabilecek bir s\u00fcper-\u00e7\u00f6z\u00fcn\u00fcrl\u00fck \u015femas\u0131 tasarlamakt\u0131r.<\/span><\/p>\n

B\u00f6yle bir s\u00fcper-\u00e7\u00f6z\u00fcn\u00fcrl\u00fck yakla\u015f\u0131m\u0131 \u015fu \u015fekildedir:<\/span><\/strong><\/p>\n

\u2022 yerel yap\u0131da bir g\u00f6r\u00fcnt\u00fcn\u00fcn hangi kom\u015fuluklar\u0131 benzerdir?<\/span>
\n\u2022 bu mahallelerin \u00f6l\u00e7ekte nas\u0131l bir ili\u015fki i\u00e7inde oldu\u011fu.<\/span><\/p>\n

Hangi mahalleler sorusuna cevap verebilmek i\u00e7in, yerel mahallelerin g\u00f6r\u00fcnt\u00fc uzay\u0131 b\u00f6l\u00fcnecektir. Daha \u00f6nce ima edildi\u011fi gibi, bu, bir\u00e7o\u011funun mevcut oldu\u011fu bir VQ algoritmas\u0131 arac\u0131l\u0131\u011f\u0131yla ger\u00e7ekle\u015ftirilir.<\/span><\/p>\n

Nas\u0131l ili\u015fkili olduklar\u0131n\u0131 belirlemek i\u00e7in, VQ’dan kaynaklanan her bir Voronoi h\u00fccresi, o k\u00fcmedeki d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc mahalleler ile onlar\u0131n homolog y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc mahalleleri aras\u0131ndaki en iyi e\u015flemeyi bulmak i\u00e7in e\u011fitilmi\u015f bir do\u011frusal \u00e7a\u011fr\u0131\u015f\u0131msal belle\u011fe (LAM) ba\u011flanacak ve b\u00f6ylece \u00f6l\u00e7ekler aras\u0131 bilgidir.<\/span><\/p>\n

Ba\u015fka bir deyi\u015fle, yapt\u0131\u011f\u0131m\u0131z varsay\u0131m, kod \u00e7izelgesi vekt\u00f6rlerinde ve LAM’lerde yer alan bilgilerin, d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc bir kom\u015fuluk ile y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc muadili aras\u0131ndaki ili\u015fkiyi (e\u015fleme) tan\u0131mlad\u0131\u011f\u0131d\u0131r. Bu nedenle, yakla\u015f\u0131m\u0131m\u0131z, k\u00f6t\u00fc konumlanm\u0131\u015f s\u00fcper\u00e7\u00f6z\u00fcn\u00fcrl\u00fck sorununa makul bir \u00e7\u00f6z\u00fcm elde etmek i\u00e7in tipik olarak gereken varsay\u0131mlar\u0131 gerektirmez.<\/span><\/p>\n

LAM’ler, g\u00f6r\u00fcnt\u00fc bilgilerini \u00f6l\u00e7ekler aras\u0131nda ili\u015fkilendiren yeniden yap\u0131land\u0131rma \u00e7ekirdekleri olarak g\u00f6r\u00fclebilir. \u00c7ekirdekleri tasarlamak i\u00e7in uyarlanabilir bir \u015fema se\u00e7iyoruz \u00e7\u00fcnk\u00fc temsili bir e\u011fitim g\u00f6r\u00fcnt\u00fcleri seti verildi\u011finde en uygun e\u015fleyicileri nas\u0131l tasarlayaca\u011f\u0131m\u0131z\u0131 biliyoruz.<\/span><\/p>\n

Ayr\u0131ca, e\u011fer yerel b\u00f6lgeler yeterince k\u00fc\u00e7\u00fckse, bilgilerin g\u00f6r\u00fcnt\u00fcler aras\u0131nda genelle\u015ftirilmesini bekliyoruz. Yeni bir g\u00f6r\u00fcnt\u00fc sunuldu\u011funda, her bir yerel b\u00f6lgeyi en iyi yeniden yap\u0131land\u0131ran \u00e7ekirdek otomatik olarak se\u00e7ilir ve yeniden yap\u0131lanma \u00e7\u0131kt\u0131da g\u00f6r\u00fcn\u00fcr.<\/span><\/p>\n

Bu metodolojinin, \u00f6rnekleme teorisine dayanan metotlardan daha iyi yeniden yap\u0131land\u0131rma sa\u011flamas\u0131 beklenebilir. Bununla birlikte, \u00f6rnekleme teorisinin evrensel karakterinden farkl\u0131 olarak, bu s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck y\u00f6ntemi, g\u00f6r\u00fcnt\u00fclerin karakterine \u00f6zg\u00fcd\u00fcr.<\/span><\/p>\n

Yani, bir g\u00f6r\u00fcnt\u00fc s\u0131n\u0131f\u0131 i\u00e7in elde edilen bazlar, ba\u015fka bir s\u0131n\u0131f yeniden olu\u015fturulurken d\u00fc\u015f\u00fck performans g\u00f6sterebilir. Bu nedenle, verilerimizin kar\u015f\u0131la\u015ft\u0131r\u0131laca\u011f\u0131 uygun modellerin olu\u015fturulmas\u0131, bir g\u00f6r\u00fcnt\u00fcn\u00fcn ba\u015far\u0131l\u0131 s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fc i\u00e7in \u00f6nemlidir.<\/span><\/p>\n

\n

Dijital g\u00f6r\u00fcnt\u00fc i\u015fleme Ders Notlar\u0131<\/span>
\nAnalog g\u00f6r\u00fcnt\u00fc i\u015fleme<\/span>
\nDijital g\u00f6r\u00fcnt\u00fc<\/a> i\u015fleme PDF<\/span>
\nG\u00f6r\u00fcnt\u00fc i\u015fleme y\u00f6ntemleri<\/span>
\nsay\u0131sal g\u00f6r\u00fcnt\u00fc i\u015fleme rafael c. gonzalez pdf<\/span>
\nG\u00f6r\u00fcnt\u00fc \u0130\u015eLEMEN\u0130N TEMELLER\u0130 Ders Notlar\u0131<\/span>
\nG\u00f6r\u00fcnt\u00fc i\u015fleme maskeleme<\/span>
\nDijital g\u00f6r\u00fcnt\u00fc nedir<\/span><\/p>\n

\n

G\u00f6r\u00fcnt\u00fc Edinme Modeli<\/span><\/strong><\/h3>\n

G\u00f6r\u00fcnt\u00fc elde etme s\u00fcreci bu b\u00f6l\u00fcmde modellenmi\u015ftir. Orijinal g\u00f6r\u00fcnt\u00fcn\u00fcn d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc bir kar\u015f\u0131l\u0131\u011f\u0131n\u0131 sentezlemek i\u00e7in bu modeli kullan\u0131yoruz. Bu modelle, d\u00fc\u015f\u00fck ve y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc mahallelerimizin destek b\u00f6lgeleri homologtur (yani, g\u00f6r\u00fcnt\u00fclenen sahnenin ayn\u0131 fiziksel b\u00f6lgesini kapsarlar).<\/span><\/p>\n

Buradaki model, bu \u015fekilde enterpolasyonlu sin c olan 128 \u00d7 128 g\u00f6r\u00fcnt\u00fcs\u00fcn\u00fc elde etmek i\u00e7in kullan\u0131ld\u0131. S\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc mimaride, d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc sentez, \u00f6l\u00e7ekler aras\u0131 bilgilerin modellenebilece\u011fi bir girdi olu\u015fturur (g\u00f6r\u00fcnt\u00fc \u00e7iftlerinden).<\/span><\/p>\n

xa(t,t1,t2) fonksiyonunun bir sens\u00f6r d\u00fczlemine \u00e7arpan s\u00fcrekli, zamanla de\u011fi\u015fen bir g\u00f6r\u00fcnt\u00fcy\u00fc temsil etmesine izin verin. Uzamsal d\u00fczlem, t1, t2 koordinat eksenleri taraf\u0131ndan referans al\u0131n\u0131r ve zaman, t de\u011fi\u015fkeni taraf\u0131ndan referans al\u0131n\u0131r. G\u00f6r\u00fcnt\u00fcleme sens\u00f6r\u00fc d\u00fczleminin, N1 \u00d7 N2 dikd\u00f6rtgen sens\u00f6r elemanlar\u0131ndan olu\u015fan bir \u0131zgara oldu\u011fu varsay\u0131l\u0131r.<\/span><\/p>\n

Bu \u00f6\u011feler, kameran\u0131n g\u00f6r\u00fc\u015f alan\u0131 i\u00e7indeki uzamsal d\u00fczlemi \u00f6rneklemeye yarar. Bu elemanlar\u0131n her birinin fiziksel boyutlar\u0131n\u0131n p1 \u00d7 p2 oldu\u011fu s\u00f6ylenir. Her eleman\u0131n \u00e7\u0131kt\u0131s\u0131, belirli bir zaman aral\u0131\u011f\u0131nda her sens\u00f6re \u00e7arpan \u0131\u015f\u0131k miktar\u0131yla orant\u0131l\u0131d\u0131r. xl[n1,n2]burada n1 =0,1,…,N1\u22121 ve n2 = 0,1,…,N2 \u2212 1 taraf\u0131ndan verilen her sens\u00f6r\u00fcn \u00e7\u0131kt\u0131s\u0131 \u015fu \u015fekilde ifade edilebilir.<\/span><\/p>\n

Daha y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc bir g\u00f6r\u00fcnt\u00fc elde etmek i\u00e7in, xl[n1,n2] elde etmede kullan\u0131lan ayn\u0131 g\u00f6r\u00fc\u015f alan\u0131n\u0131 kapsayan daha ince bir sens\u00f6r \u0131zgaras\u0131 kullan\u0131lmal\u0131d\u0131r. Her bir uzamsal boyuttaki \u00e7\u00f6z\u00fcn\u00fcrl\u00fc\u011f\u00fcn ilgili uzamsal boyutlarda G1 ve G2 fakt\u00f6r\u00fc kadar artmas\u0131na izin verin.\u00a0<\/span><\/p>\n

Sens\u00f6r elemanlar\u0131n\u0131n fiziksel boyutu art\u0131k p1 \u00d7 p2 alan birimi olur. Y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc G1 G2 imageisthengivenbyxh[m1,m2],buradam1 =0,1,…,M1\u22121vem2 =0,1,…,M2\u22121 ve Mi = GiNi(i = 1,2) . Y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc g\u00f6r\u00fcnt\u00fc i\u00e7in M1 \u00d7 M2 sens\u00f6r \u00f6\u011felerinin her birinin \u00e7\u0131kt\u0131s\u0131 a\u00e7\u0131klanabilir.<\/span><\/p>\n

G\u00f6r\u00fcnt\u00fcdeki her piksel i\u00e7in ortalama yo\u011funluk de\u011ferini korumak amac\u0131yla zaman i\u00e7indeki entegrasyon s\u0131n\u0131rlar\u0131n\u0131n bir zaman biriminden G1G2 zaman birimlerine uzat\u0131ld\u0131\u011f\u0131na dikkat edin.<\/span><\/p>\n

S\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck i\u015flemi, d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc g\u00f6r\u00fcnt\u00fc xl[n1,n2]’den y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc xh[m1,m2] g\u00f6r\u00fcnt\u00fcs\u00fcn\u00fc tahmin etmektir. xh[m1,m2]’den xl[n1,n2] elde etme i\u015fleminin verildi\u011fine dikkat \u00e7ekilebilir.<\/span><\/p>\n

Yukar\u0131daki denklemdeki desimasyon modeli, y\u00fcksek \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc g\u00f6r\u00fcnt\u00fcdeki G1 \u00d7 G2 \u00f6rt\u00fc\u015fmeyen piksel kom\u015fuluklar\u0131n\u0131n piksellerinin ortalamas\u0131n\u0131 alarak d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc bir g\u00f6r\u00fcnt\u00fc \u00fcretir.<\/span><\/p>\n

\u00c7ekirdek Tabanl\u0131 Yakla\u015f\u0131mlar\u0131 \u0130li\u015fkilendirme<\/span><\/strong><\/h3>\n

Bu b\u00f6l\u00fcm, g\u00f6r\u00fcnt\u00fclerin b\u00fcy\u00fct\u00fclmesi i\u00e7in \u00e7ekirdek tabanl\u0131 bi\u00e7imcilikleri tan\u0131tmaktad\u0131r. Geleneksel b\u00fcy\u00fctme yakla\u015f\u0131mlar\u0131, bir g\u00f6r\u00fcnt\u00fcn\u00fcn \u00f6rnek yo\u011funlu\u011funu art\u0131rmak i\u00e7in tek bir \u00e7ekirdek kullan\u0131r ve \u00f6rnekler aras\u0131nda enterpolasyon yapar. Burada sunulan s\u00fcper \u00e7\u00f6z\u00fcn\u00fcrl\u00fck metodolojisi, bir \u00e7ekirdek ailesinin kullan\u0131m\u0131yla ilgilidir. Her \u00e7ekirdek, \u00f6l\u00e7ekler aras\u0131nda bir g\u00f6r\u00fcnt\u00fcn\u00fcn belirli bilgilerine g\u00f6re uyarlanm\u0131\u015ft\u0131r.<\/span><\/p>\n

Tek \u00c7ekirdek<\/span><\/strong><\/p>\n

B\u00fcy\u00fct\u00fclm\u00fc\u015f bir g\u00f6r\u00fcnt\u00fc, d\u00fc\u015f\u00fck \u00e7\u00f6z\u00fcn\u00fcrl\u00fckl\u00fc bir g\u00f6r\u00fcnt\u00fcn\u00fcn xl[n1,n2] \u00f6rneklerini geni\u015fleterek ve \u00f6rneklenmi\u015f bir enterpolasyonla evri\u015fim yaparak elde edilebilir. G1 \u00d7 G2’lik bir geni\u015fletme oran\u0131 i\u00e7in, burada G1 , G2 1’den b\u00fcy\u00fck tam say\u0131lard\u0131r, geni\u015fletilmi\u015f g\u00f6r\u00fcnt\u00fc de verilir.<\/span><\/p>\n

Bu enterpolasyon bi\u00e7imi, g\u00f6r\u00fcnt\u00fcy\u00fc ba\u015ftan sona benzer \u015fekilde i\u015fleyen do\u011frusal bir filtrelemedir (yani, Shannon \u00f6rnekleme teorisinde oldu\u011fu gibi, enterpolasyonlu noktalar\u0131n belirlenmesinde g\u00f6r\u00fcnt\u00fc \u00f6rneklerinin ayn\u0131 do\u011frusal kombinasyonunu da kullan\u0131r).<\/span><\/p>\n

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