Comparison of image retargeting algorithms with Seam Carving method

Basit Öğe Kaydı
dc.authorid0000-0001-7301-9085
dc.contributor.advisorEkin, Emineen_US
dc.contributor.authorMiroğlu, Taylanen_US
dc.contributor.otherIşık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans Programıen_US
dc.date.accessioned2023-08-07T12:20:54Z
dc.date.available2023-08-07T12:20:54Z
dc.date.issued2023-04-25
dc.departmentIşık Üniversitesi, Lisansüstü Eğitim Enstitüsü, Bilgisayar Mühendisliği Yüksek Lisans Programıen_US
dc.descriptionText in English ; Abstract: English and Turkishen_US
dc.descriptionIncludes bibliographical references (leaves 45-46)en_US
dc.descriptionx, 47 leavesen_US
dc.description.abstractThe rise of social media has made sharing photos and pictures more important than ever, both for personal and marketing purposes. This situation also caused the problem of converting the photos taken with the camera in a square format, where the width is higher than the height. To address this need, a recent study explored the use of the Seam Carving method to convert images to a square format while preserving their essential parts. The study compared two algorithms, Greedy and Dijkstra, in terms of processing time and consistency using a supervised image. The consistency comparison was carried out on five images, three of which were obtained from NRID, and two were created for the study. The five images were used to calculate the average consistency of the Dijkstra algorithm. In addition, 23 more images from NRID were used to compute the average consistency of the Greedy algorithm, resulting in a total of 28 images used in the analysis. The results showed that the Greedy algorithm had an average consistency that was 6.55% higher than the Dijkstra algorithm based on the five images. Furthermore, the Dijkstra algorithm took an average of 2,347% longer to process than the Greedy algorithm. The implications of these findings are significant for social media users and marketers alike. The Greedy algorithm can help maintain the essential elements of an image while making it suitable for different social media platforms. The study also highlights the importance of considering processing time when choosing an algorithm to use. Overall, this research demonstrates the potential of the Seam Carving method and provides valuable insights into the choice of algorithm for image manipulation.en_US
dc.description.abstractSosyal medyanın yükselişi, kişisel ve pazarlama amaçları için fotoğraf ve resim paylaşımını daha da önemli hale getirdi. Bu durum aynı zamanda, kamera ile çekilen ve genişliği yüksekliğinden daha fazla olan fotoğrafların kare formata dönüştürülmesi sorununu da beraberinde getirdi. Bu ihtiyacı karşılamak için son zamanlarda bir çalışma, resimleri özgün parçalarını koruyarak kare formata dönüştürmek için Seam Carving yönteminin kullanımını inceledi. Bu çalışmada, süpervize edilmiş bir görüntü üzerinde hem işlem süresi hem de tutarlılık açısından Greedy yaklaşım ve Dijkstra algoritması olmak üzere iki algoritma karşılaştırdı. Bu araştırmadaki tutarlılık karşılaştırmasında beş görüntü kullanıldı; üç tanesi NRID'den elde edilen ve iki tanesi bu çalışma için özel olarak oluşturulan beş görüntü üzerinde yapıldı. Beş görüntü, Dijkstra algoritmasının ortalama tutarlılığını hesaplamak için kullanıldı. Bunun yanı sıra, NRID'den 23 tane daha görüntü, Greedy algoritmasının ortalama tutarlılığını hesaplamak için kullanıldı. Bu araştırmanın analizinde toplamda 28 görüntü kullanıldı. Sonuçlar, beş farklı görüntüye dayanarak Greedy algoritmasının ortalama tutarlılığının Dijkstra algoritmasından %6,55 daha yüksek olduğunu gösterdi. Bunun yanı sıra, Dijkstra algoritmasına ait işlem süresinin Greedy algoritmasından %2.347 daha uzun sürdüğü ortaya çıktı. Bu bulguların sosyal medya kullanıcıları ve pazarlamacılar için önemli sonuçları vardır. Greedy algoritması, bir görüntünün temel öğelerini koruyarak farklı sosyal medya platformlarına uygun hale getirmeye yardımcı olabilir. Bu çalışma, görüntü yeniden boyutlandırma yöntemlerinden olan Seam Carving yönteminde algoritma seçiminde işlem süresinin dikkate alınmasının önemini vurgulamaktadır. Genel olarak, bu araştırma, Seam Carving yönteminin potansiyelini göstermektedir ve görüntü manipülasyonu için algoritma seçimi konusunda değerli bilgiler sağlamaktadır.en_US
dc.description.tableofcontentsApplication of Image Retargeting Algorithmsen_US
dc.description.tableofcontentsAuxiliaries, Libraries and Languageen_US
dc.description.tableofcontentsStepsen_US
dc.description.tableofcontentsPreparing the Dataseten_US
dc.description.tableofcontents3-Dimensional Colored Image to 2-Dimensional Grayscale Imageen_US
dc.description.tableofcontentsLaplacian Transformen_US
dc.description.tableofcontentsFinding The Lowest Energy Points – The Shortest Pathen_US
dc.description.tableofcontentsFirst-level Greedy Approachen_US
dc.description.tableofcontentsDijkstra Algorithmen_US
dc.description.tableofcontentsMarking Red – Removing Pixelsen_US
dc.description.tableofcontentsProcess Sequence for First-level Greedy Approachen_US
dc.description.tableofcontentsProcess Sequence for Dijkstra Algorithmen_US
dc.description.tableofcontentsIntersection over Union (IoU)en_US
dc.description.tableofcontentsMean Results and Comparisonen_US
dc.description.tableofcontentsVisual Comparison with Other Methodsen_US
dc.description.tableofcontentsCompared Laplacian filtersen_US
dc.description.tableofcontentsComparison of Optimized and Classical Greedy Approachen_US
dc.description.tableofcontentsThe logic of Dijkstra Algorithmen_US
dc.description.tableofcontentsIoU of first-level greedy approach photoen_US
dc.description.tableofcontentsIoU of Dijkstra photoen_US
dc.description.tableofcontentsIoU of first-level greedy approach imageen_US
dc.description.tableofcontentsIoU of Dijkstra imageen_US
dc.description.tableofcontentsIoU of first-level greedy approach image in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsIoU of Dijkstra image in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsIoU of first-level greedy approach image in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsIoU of Dijkstra image in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsIoU of first-level greedy approach image in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsIoU of Dijkstra image in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsIoU of greedy approachen_US
dc.description.tableofcontentsIoU of Dijkstraen_US
dc.description.tableofcontentsMean accuracy (IoU) of the all the imageen_US
dc.description.tableofcontentsRun time and accuracy comparison of greedy approach and Dijkstra algorithm on photoen_US
dc.description.tableofcontentsRun time and accuracy comparison of greedy approach and Dijkstra algorithm on imageen_US
dc.description.tableofcontentsRun time and accuracy comparison of greedy approach (v1 & v2) and Dijkstra algorithm on the image in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsRun time and accuracy comparison of greedy approach (v1 & v2) and Dijkstra algorithm on the image in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsRun time and accuracy comparison of greedy approach (v1 & v2) and Dijkstra algorithm on the image in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsRun time and accuracy comparison of greedy approach (v2) and Dijkstra algorithm on the mean of all the imagesen_US
dc.description.tableofcontentsThe 2-D Laplacian function (researchgate.net)en_US
dc.description.tableofcontentsThe original and Laplacian filtered photosen_US
dc.description.tableofcontentsThe original and Laplacian filtered imagesen_US
dc.description.tableofcontentsThe original and Laplacian filtered images in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsThe original and Laplacian filtered images in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsThe original and Laplacian filtered images in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsGreedy approachen_US
dc.description.tableofcontentsGreedy approach (Next Step)en_US
dc.description.tableofcontentsThe logic of Dijkstra Algorithmen_US
dc.description.tableofcontentsRed marks on the original and Laplacian filtered photos with using greedy approachen_US
dc.description.tableofcontentsRed marks on the original and Laplacian filtered images with using greedy approachen_US
dc.description.tableofcontentsRed marks on the original and Laplacian filtered images in NRID (ours_11_aaa) with using greedy approachen_US
dc.description.tableofcontentsWhy algorithm needs adjusting process before marking pixels –Step 1en_US
dc.description.tableofcontentsWhy algorithm needs adjusting process before marking pixels –Step 2en_US
dc.description.tableofcontentsWhy algorithm needs adjusting process before marking pixels –Step 3en_US
dc.description.tableofcontentsRed marks on the original and Laplacian filtered photos with using Dijkstra algorithmen_US
dc.description.tableofcontentsRed marks on the original and Laplacian filtered images with using Dijkstra algorithmen_US
dc.description.tableofcontentsRed marks on the original and Laplacian filtered images in NRID (ours_11_aaa) with using Dijkstra algorithmen_US
dc.description.tableofcontentsFlowchart of the Algorithmen_US
dc.description.tableofcontentsThe original and retargeted photo with greedy approachen_US
dc.description.tableofcontentsThe original and retargeted photo with Dijkstra algorithmen_US
dc.description.tableofcontentsThe original and retargeted image with greedy approachen_US
dc.description.tableofcontentsThe original and retargeted image with Dijkstra algorithmen_US
dc.description.tableofcontentsThe original and retargeted image with Greedy algorithm in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsThe original and retargeted image with Dijkstra algorithm in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsThe original and retargeted image with Greedy algorithm in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsThe original and retargeted image with Dijkstra algorithm in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsThe original and retargeted image with Greedy algorithm in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsThe original and retargeted image with Dijkstra algorithm in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsSupervised photoen_US
dc.description.tableofcontentsSupervised imageen_US
dc.description.tableofcontentsSupervised image in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsSupervised image in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsSupervised image in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsCreating Polylines in AutoCAD for Getting Coordinatesen_US
dc.description.tableofcontentsVisual representation of IoU of photoen_US
dc.description.tableofcontentsVisual representation of IoU of imageen_US
dc.description.tableofcontentsVisual representation of IoU of image in NRID (ours_11_aaa)en_US
dc.description.tableofcontentsVisual representation of IoU of image in NRID (ours_14_aaa)en_US
dc.description.tableofcontentsVisual representation of IoU of image in NRID (ours_16_aaa)en_US
dc.description.tableofcontentsCropped photo and imageen_US
dc.description.tableofcontentsStretched photo and imageen_US
dc.description.tableofcontentsCropped vs Stretched vs Greedy seam carving of photoen_US
dc.description.tableofcontentsCropped vs Stretched vs Greedy seam carving of imageen_US
dc.description.tableofcontentsCropped vs Stretched vs Dijkstra seam carving of photoen_US
dc.description.tableofcontentsCropped vs Stretched vs Dijkstra seam carving of imageen_US
dc.description.tableofcontentsCropped vs Stretched vs Greedy seam carving of photoen_US
dc.description.tableofcontentsCropped vs Stretched vs Dijkstra seam carving of photoen_US
dc.description.tableofcontentsCropped vs Stretched vs Greedy seam carving of imageen_US
dc.description.tableofcontentsCropped vs Stretched vs Dijkstra seam carving of imageen_US
dc.description.tableofcontentsGreedy vs Dijkstra seam carving of photoen_US
dc.description.tableofcontentsGreedy vs Dijkstra seam carving of imageen_US
dc.identifier.citationMiroğlu, T. (2023). Comparison of image retargeting algorithms with Seam Carving method. İstanbul: Işık Üniversitesi Lisansüstü Eğitim Enstitüsü.en_US
dc.identifier.urihttps://hdl.handle.net/11729/5675
dc.institutionauthorMiroğlu, Taylanen_US
dc.institutionauthorid0000-0001-7301-9085
dc.language.isoenen_US
dc.publisherIşık Üniversitesien_US
dc.relation.publicationcategoryTezen_US
dc.rightsinfo:eu-repo/semantics/openAccessen_US
dc.rightsAttribution-NonCommercial-NoDerivs 3.0 United States*
dc.rights.urihttp://creativecommons.org/licenses/by-nc-nd/3.0/us/*
dc.subjectSeam Carvingen_US
dc.subjectDijkstraen_US
dc.subjectGreedyen_US
dc.subjectImage retargetingen_US
dc.subjectImage resizingen_US
dc.subjectShortest pathen_US
dc.subjectResim yeniden hedeflemeen_US
dc.subjectResim boyutlandırmaen_US
dc.subjectEn kısa yolen_US
dc.subject.lccT357 .M57 C66 2023
dc.subject.lcshSeam carving.en_US
dc.subject.lcshDijkstra.en_US
dc.subject.lcshGreedy.en_US
dc.subject.lcshImage retargeting.en_US
dc.subject.lcshImage resizing.en_US
dc.subject.lcshShortest path.en_US
dc.titleComparison of image retargeting algorithms with Seam Carving methoden_US
dc.title.alternativeSeam Carving yöntemi ile görüntü yeniden hedefleme algoritmalarının karşılaştırılmasıen_US
dc.typeMaster Thesisen_US
dspace.entity.typePublication

Dosyalar

Orijinal paket
Listeleniyor 1 - 1 / 1
Küçük Resim
İsim:
Comparison_of_image_retargeting_algorithms_with_Seam_Carving_method.pdf
Boyut:
3.48 MB
Biçim:
Adobe Portable Document Format
Açıklama:
MasterThesis
İndir
Lisans paketi
Listeleniyor 1 - 1 / 1
Küçük Resim Yok
İsim:
license.txt
Boyut:
1.44 KB
Biçim:
Item-specific license agreed upon to submission
Açıklama:
İndir

Koleksiyon

Lisansüstü Eğitim Enstitüsü Tez Koleksiyonu