20.500.12556/RUL-73477 Nekateri postopki za izboljšanje in poravnavanje podvodnih slik Some methods for enhancement and registration of underwater images V pričujočem delu so bile preizkušene nekatere metode obdelav podvodnih slik, z namenom vizualne izboljšave in izgradnje panoramskih slik. Preizkušene so bile nekatere metode izboljšanja slik s preslikavo sivin, kot so gama preslikave, metoda ravnanja histogramov in metoda adaptivnega ravnanja histogramov. Preizkušene so bile tudi nekatere metode obnavljanja slik, kot so na primer ostrenje z megljenjem, slepa dekonvolucija in Wienerjevo filtriranje. Sledilo je iskanje značilk na podlagi SIFT algoritma in nato s pomočjo njih, še izračun ustreznih transformacijskih matrik, ki jih potrebujemo za poravnavo slik. Drugi del naloge je zajemal preizkušanje raznih metod izgradnje panoramskih slik, kot so metode lepljenja slik, izgradnja slike s pomočjo medianinega filtriranja kupčka poravnanih slik in metoda, kjer posamezne slikovne elemente prepišemo iz tiste slike v kupčku poravnanih slik, katere center je najbližji temu slikovnemu elementu. Celotna realizacija naloge je bila preizkušena in izvedena v programskem okolju MATLAB. Uspešnost izboljšanja in obnavljanja slik smo določili z očnim primerjanjem slik pred in po obdelavi. Med preizkušenimi metodami izboljšanja in obnavljanja slik sta se kot najbolj primerni izkazali metodi adaptivnega ravnanja histogramov in ostrenja z megljenjem. Med preizkušenimi metodami za izgradnjo panoramskih slik, pa je bila najbolj primerna zadnja izmed prej naštetih metod. The thesis addresses some known methods of underwater image processing with the goal of visualy improving images and creating a larger panoramic image. The first aim of the thesis is to use known and promising methods for image enhancement and image restoration and apply them to underwater video sequences of the sea floor. Image enhancement methods that were being tested were gamma correction, histogram equalization and adaptive histogram equalization. Image restoration methods that were being tested were unsharp masking, blind deconvolution and wiener filtering. The second aim of the thesis is to produce a panoramic image of a better quality than the individual images by geometric alignment of images in a time sequence and by combining overlapping images together. The first step was to match neighboring frames by extracting features using a SIFT algorithm and then using those feature pairs, to estimate geometric transformation matrices, that we need for alignment of frames. Tested methods for creating panoramic images were blending, median filtering and a method, that built a panoramic image by selecting pixels of those images, whose centers were closest to the location of that pixel. The whole thesis was realized and tested in MATLAB. We selected the most appropriate method of image enhancement and restoration by visually comparing images before and after processing. Of all the tested methods of image enhancement and restoration, adaptive histogram equalization and unsharp masking were picked as the most appropriate one. Of all the tested methods of panoramic image creation, the one that based on calculating distances was picked as the most appropriate one. podvodno snemanje izboljšava slik restavracija slik ravnanje histogramov panoramske slike značilke geometrijske preslikave underwater imaging image enhancement image restoration histogram equalization panoramic images features geometric transforms true false false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2015-11-18 13:29:20 2015-11-18 13:29:20 2022-08-01 17:57:45 0000-00-00 00:00:00 2015 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 1970-01-01 31081 Goljat_Rok_-_Nekateri_postopki_za_izboljsanje_in_poravnavanje_podvodnih_slik.pdf Goljat_Rok_-_Nekateri_postopki_za_izboljsanje_in_poravnavanje_podvodnih_slik.pdf 1 C35D3B5FA66C1A53FEFE726EDD72A865 578d7f7b7f40d547d43feaeadb37c85620f2da199813bebf87eb85b6cd8f51c9 64dbeb93-a1ad-11eb-a523-00155dcfd717 20.500.12556/rul/6e3926b9-5e21-48ed-a8dd-216749314644 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=73515 Fakulteta za elektrotehniko 0 0 0