Procesiranje slik s programskim jezikom Halide

DUKIČ, ROK (author), Skočaj, Danijel (mentor)

V diplomskem delu je predstavljen novejši programski jezik za obdelavo slik Halide in njegova že uveljavljena alternativa OpenCV ter njuna primerjava. Ta obsega čas, potreben za izvedbo funkcij in reševanje enakih problemov, ter dolžino implementacij v številu vrstic. Problemi se stopnjujejo po kompleksnosti. Ti obsegajo implementacije za morfološke operacije in operacijo za zaznavanje objekta na sliki z uporabo vzorcev. Vse operacije bodo opisane in implementirane na štiri načine. Prvi bo izdelan v jeziku C++ in bo za predstavitev slik uporabljal objekte iz OpenCV, drugi bo uporabljal funkcijo iz OpenCV, tretji in četrti pa bosta izdelana v programskem jeziku Halide. Razlika bo v tem, da eden ne bo uporabljal razvrščanja, zato bo bolj počasen. V drugi polovici diplomskega dela so rezultati predstavljeni s tabelami, na koncu pa je še obrazložitev rezultatov.

• Language: Slovenian
• Keywords: Halide, OpenCV, morfološke operacije, ujemanje vzorca, obdelava slik, računalniški vid
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2017
• PID: 20.500.12556/RUL-89083
• Publication date in RUL: 14.02.2017

Secondary language

• Language: English
• Title: Processing images with programming language Halide

Abstract

The thesis contains a presentation of a recently created programming language Halide and its comparison to an already established image processing library OpenCV. We compare the execution times of the implementations with the same functionality and their length (in terms of number of lines). The implementations consist of morphological operations and template matching. Operations are implemented in four versions. The first version is made in C++ and only uses OpenCV’s objects. The second version uses the OpenCV’s function. The third and fourth version are written in Halide language. The difference between them is in the optimization. One of them uses scheduling while the other does not. The second part of the thesis consists of the results which are shown in tables and the analysis of the results.

• Keywords: Halide, OpenCV, morphology, template matching, image processing, machine vision