izpis_h1_title_alt

Kompresija slik z uporabo konvolucijskih nevronskih mrež
Selimović, Alena (Author), Hladnik, Aleš (Mentor) More about this mentor... This link opens in a new window

.pdfPDF - Presentation file, Download (4,06 MB)

Abstract
Konvolucijske nevronske mreže so se izkazale kot uspešne pri marsikateri nalogi s področja slikovnega procesiranja in računalniškega vida. Namen diplomskega dela je bila izdelava konvolucijske nevronske mreže, ki omogoča lokalizacijo interesnih območij na sliki. Lokalizacija je izražena kot pripadajoči toplotni zemljevid, ki z uporabo različnih hladnih in toplih barv opredeli pomembnost posameznih območij na sliki. Za izgradnjo nevronske mreže smo uporabili predhodno učeni model VGG16, katerega arhitekturo smo prilagodili glede na specifične zahteve zadane naloge. Na osnovi toplotnega zemljevida smo za vsak piksel vhodne slike pridobili vrednost, ki nakazuje njegovo pomembnost. Dobljene vrednosti smo diskretizirali v več nivojev in vsak nivo kodirali z različnim faktorjem kakovosti Q tako, da smo pomembnejše nivoje kodirali z višjo bitno stopnjo, manj pomembne pa z nižjo bitno stopnjo, dobljeno sliko pa smo nato stisnili še s standardnim JPEG algoritmom. Rekonstruirane slike smo primerjali s slikami, ki so bile stisnjene in nato dekodirane s standardno JPEG kompresijo pri Q faktorjih 30, 50 in 70. Za objektivno primerjavo in vrednotenje kakovosti rekonstruiranih slik smo uporabili metodo srednje kvadratne napake (MSE), metodo maksimalnega razmerja med signalom in šumom (PSNR), indeks strukturne podobnosti (SSIM) in metodo, ki indeks strukturne podobnosti izračuna v več iteracijah (MS-SSIM). Rezultati so pokazali, da metoda kompresije, ki temelji na uporabi toplotnih zemljevidov, pri enakih ali celo manjših velikostih datotek omogoča manjšo napako (MSE) in višje vrednosti PSNR rekonstruiranih slik v primerjavi s standardno JPEG kompresijo. Višji so tudi izračunani indeksi SSIM in MS-SSIM, kar pomeni, da se kakovost rekonstruiranih slik tudi vizualno bolj ujema s človeško oceno kot kakovost običajnih JPEG slik.

Language:Slovenian
Keywords:konvolucijske nevronske mreže, toplotni zemljevidi, interesna območja, kompresija slik, JPEG
Work type:Bachelor thesis/paper (mb11)
Organization:NTF - Faculty of Natural Sciences and Engineering
Year:2018
Views:477
Downloads:210
Metadata:XML RDF-CHPDL DC-XML DC-RDF
 
Average score:(0 votes)
Your score:Voting is allowed only to logged in users.
:
Share:AddThis
AddThis uses cookies that require your consent. Edit consent...

Secondary language

Language:English
Title:Image Compression Using Convolutional Neural Networks
Abstract:
Convolutional neural networks have proved to be successful in various image processing and computer vision tasks. The aim of the diploma thesis was to create a convolutional neural network that enables the localization of regions of interest (ROIs) in an image. The localization is expressed as a corresponding heatmap, which uses a colour scheme, consisting of a range of cold and warm colours, to identify the importance of individual image regions. The neural network was built using the pretrained VGG16 model, the architecture of which was modified in accordance with the specific demands of the chosen task. The heatmap provided a saliency value for each pixel of the given image. These values were then discretized into multiple levels and each level was encoded with a different quality factor Q by encoding the more important levels at a higher bitrate, and the less important levels at a lower bitrate; the obtained image was then encoded further using the standard JPEG algorithm. The reconstructed images were compared to images that had been compressed and then decoded using standard JPEG compression at Q factors of 30, 50 and 70. For an objective comparison and evaluation of the quality of the reconstructed images the mean squared error (MSE), peak signal-to-noise ratio (PSNR), structural similarity index (SSIM) and multi-scale structural similarity index (MS-SSIM) were used. The results showed that in files of the same or even smaller size the compression method, based on the obtained heatmaps, allows for a lower error (MSE) and higher PSNR values of reconstructed images in comparison with standard JPEG images. The calculated SSIM and MS-SSIM indexes were higher as well, which shows that the visual quality of the reconstructed images matches human evaluation better than the visual quality of standard JPEG images.

Keywords:convolutional neural networks, heatmaps, regions of interest, image compression, JPEG

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Comments

Leave comment

You have to log in to leave a comment.

Comments (0)
0 - 0 / 0
 
There are no comments!

Back