Spreminjanje velikosti slik glede na vsebino

Matkovič, Tilen (author), Šajn, Luka (mentor)

Cilj diplomskega dela je realizirati in napisati program, ki vhodni sliki spremeni velikost, tako da se čimbolj ohranijo pomembni objekti, ter opisati nekaj algoritmov, s katerimi se to lahko doseže. V uvodu je opisan problem, ki ga v diplomi rešujemo. Drugi del se nanaša na nekaj možnih rešitev, s katerimi lahko rešimo naš problem. To so OCR, zaznavanje obrazov, ločevanje ključnih značilnic in rezanje šivov. Vsak od teh je podrobno opisan. V naslednjem poglavju je opisan pristop k problemu. Začnemo z opisom orodij, ki smo jih uporabili pri izdelavi programa, in algoritmom, ki smo ga realizirali kot mešanico algoritmov, opisanih v drugem poglavju (ločevanje ključnih značilnic in rezanje šivov). Četrti del vsebuje primere delovanja našega programa. Rezultati so slike, ki vsebujejo samo ključne oz. pomembne objekte. Pri naslednjem delu je opisano delovanje zaznavanja obrazov, ki bi lahko prišlo v poštev k našemu algoritmu. Na samem koncu so opisani sklep, ugotovitve in možnosti nadgradnje algoritma.

• Language: Slovenian
• Keywords: OpenCV, računalniški vid, rezanje šivov, OCR, ločevanje značilnic, spreminjanje razmerja slik
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2015
• PID: 20.500.12556/RUL-72137
• Publication date in RUL: 04.09.2015

Secondary language

• Language: English
• Title: Content aware image resizing

Abstract

The aim of the thesis is to realize and write a program that resizes the input image to preserve important objects and describe some algorithms by which this can be achieved. The introduction describes the problem. The second part relates to some possible solutions that could solve the problem. These are OCR, face detection, salient feature extraction and seam carving. Each of these is described in detail. The next chapter describes the approach to the problem. We begin with descriptions of the tools that were used in the production of the algorithm, which was realized as a mixture of algorithms described in the second chapter (salient feature extraction and seam carving). The fourth part contains examples of how our program works. The results are images that only contain the important objects. The next section describes the operation of face detection, which we could realize in our algorithm. At the very end of the thesis we described the findings and possible upgrades of the algorithm.

• Keywords: OpenCV, computer vision, seam carving, OCR, feature extraction, image resizing