Добірка наукової літератури з теми "Image enhancement"

<p>This Master Thesis has been conducted at the National Laboratory of Forensic Science (SKL) in Linköping. When images that are to be analyzed at SKL, presenting an interesting object, are of bad quality there may be a need to enhance them. If several images with the object are available, the total amount of information can be used in order to estimate one single enhanced image. A program to do this has been developed by studying methods for image registration and high resolution image estimation. Tests of important parts of the procedure have been conducted. The final results are satisfying and the key to a good high resolution image seems to be the precision of the image registration. Improvements of this part may lead to even better results. More suggestions for further improvementshave been proposed.</p><br><p>Detta examensarbete har utförts på uppdrag av Statens Kriminaltekniska Laboratorium (SKL) i Linköping. Då bilder av ett intressant objekt som ska analyseras på SKL ibland är av dålig kvalitet finns det behov av att förbättra dessa. Om ett flertal bilder på objektet finns tillgängliga kan den totala informationen fråndessa användas för att skatta en enda förbättrad bild. Ett program för att göra detta har utvecklats genom studier av metoder för bildregistrering och skapande av högupplöst bild. Tester av viktiga delar i proceduren har genomförts. De slutgiltiga resultaten är goda och nyckeln till en bra högupplöst bild verkar ligga i precisionen för bildregistreringen. Genom att förbättra denna del kan troligtvis ännu bättre resultat fås. Även andra förslag till förbättringar har lagts fram.</p>

This research presents a framework for enhancing and analyzing time-sequenced mammographic images for detection of cancerous tissue, specifically designed to assist radiologists and physicians with the detection of breast cancer. By using computer aided diagnosis (CAD) systems as a tool to help in the detection of breast cancer in computed tomography (CT) mammography images, previous CT mammography images will enhance the interpretation of the next series of images. The first stage of this dissertation applies image subtraction to images from the same patient over time. Image types are defined as temporal subtraction, dual-energy subtraction, and Digital Database for Screening Mammography (DDSM). Image enhancement begins by applying image registration and subtraction using Matlab 2012a registration for temporal images and dual-energy subtraction for dual-energy images. DDSM images require no registration or subtraction as they are used for baseline analysis. The image data are from three different sources and all images had been annotated by radiologists for each image type using an image mask to identify malignant and benign. The second stage involved the examination of four different thresholding techniques. The amplitude thresholding method manipulates objects and backgrounds in such a way that object and background pixels have grey levels grouped into two dominant and different modes. In these cases, it was possible to extract the objects from the background using a threshold that separates the modes. The local thresholding introduced posed no restrictions on region shape or size, because it maximized edge features by thresholding local regions separately. The overall histogram analysis showed minima and maxima of the image and provided four feature types--mean, variance, skewness, and kurtosis. K-means clustering provided sequential splitting, initially performing dynamic splits. These dynamic splits were then further split into smaller, more variant regions until the regions of interest were isolated. Regional-growing methods used recursive splitting to partition the image top-down by using the average brightness of a region. Each thresholding method was applied to each of the three image types. In the final stage, the training set and test set were derived by applying the four thresholding methods on each of the three image types. This was accomplished by running Matlab 2012a grey-level, co-occurrence matrix (GLCM) and utilizing 21 target feature types, which were obtained from the Matlab function texture features. An additional four feature types were obtained from the state of the histogram-based features types. These 25 feature types were applied to each of the two classifications malignant and benign. WEKA 3.6.10 was used along with classifier J48 and cross-validation 10 fold to find the precision, recall, and f-measure values. Best results were obtained from these two combinations: temporal subtraction with amplitude thresholding, and temporal subtraction with regional-growing thresholding. To summarize, the researcher's contribution was to assess the effectiveness of various thresholding methods in the context of a three-stage approach, to help radiologists find cancerous tissue lesions in CT and MRI mammography images.

In this thesis, image dynamic range enhancement methods are studied in order to solve the problem of representing high dynamic range scenes with low dynamic range images. For this purpose, two main image dynamic range enhancement methods, which are high dynamic range imaging and exposure fusion, are studied. More detailed analysis of exposure fusion algorithms are carried out because the whole enhancement process in the exposure fusion is performed in low dynamic range, and they do not need any prior information about input images. In order to evaluate the performances of exposure fusion algorithms, both objective and subjective quality metrics are used. Moreover, the correlation between the objective quality metrics and subjective ratings is studied in the experiments.

Every picture tells a story. Images are one of the most dominant types of media, uploaded an average in billions every single day and in hundreds of billions on an annual basis. Artifacts depicting visual perception like photographs and other two-dimensional pictures are distributed through the growing number of image-sharing websites. Consequently, a thriving interest in understanding the whole image or objects depicted in it, its style or the emotions a picture might evoke, together with all the other image properties, became increasingly represented in research practice. This research focuses on the problem of automatically enhancing memorability of an image. Recent works in Computer Vision and Multimedia have shown that intrinsic image properties like memorability can be automatically inferred by exploiting powerful deep learning models. This research advances the state of the art in this area by addressing a novel and more challenging issue: “Can we transform an arbitrary input image and make it more memorable?”. To state this question properly one requires the existence of memorability measures. Methods for automatically increasing image memorability would have an impact in many application fields, such as education, gaming or advertising. To tackle the problem, we introduce an approach inspired by editing-by-applying-filters paradigm, adopted in photo editing applications like Instagram and Prisma. Users of the two apps generally have to go through the available filters before finding the desired solution which is turning the editing process into a resource- and time-consuming task. In the work conducted for the purpose of this thesis, we reverse the process: given an input image, we propose to automatically retrieve a set of “style seeds”, i.e., a set of style images which, applied to the input image through a neural style transfer algorithm, provide the highest increase in memorability. As a result, we demonstrate that it is possible to automatically retrieve the best style seeds for a given image, thus, remarkably reducing the number of human attempts needed to find a good match. Furthermore, we show the effectiveness of the proposed approach with experiments on the publicly available LaMem dataset, performing both a quantitative evaluation and a user study. To demonstrate the flexibility of the proposed framework, we also analyze the impact of different implementation choices, such as using different state of the art neural style transfer methods. Finally, we show several qualitative results to provide additional insights on the link between image style and memorability. This approach arises from recent advances in the field of image synthesis and adopts a deep architecture for generating a memorable picture from a given input image and a style seed. Importantly, to automatically select the best style, also relying on deep models, a novel learning-based solution is proposed. The experimental evaluation, conducted on publicly available benchmarks, demonstrates the effectiveness of the proposed approach for generating memorable images through automatic style seed selection.<br>Ogni immagine racconta una storia. Le immagini sono uno dei tipi di media più dominanti. Ogni giorno ne vengono caricate miliardi, per un totale di centinaia di miliardi in media ogni anno. Gli artefatti che rappresentano la percezione visiva come fotografie e altre immagini bidimensionali sono distribuiti attraverso un numero crescente di siti web di condivisione di immagini. Di conseguenza, un crescente interesse nel comprendere l'intera immagine o gli oggetti in essa raffigurati, il suo stile o le emozioni che un'immagine potrebbe evocare, insieme a tutte le altre proprietà dell'immagine, è diventato sempre più diffuso nella pratica di ricerca. Questa ricerca si concentra sul problema del miglioramento automatico della memorabilità di un'immagine. Recenti lavori in Computer Vision e Multimedia hanno dimostrato che proprietà intrinseche dell'immagine come la memorabilità possono essere dedotte automaticamente sfruttando potenti modelli di deep learning. Questa ricerca fa avanzare lo stato dell'arte in questo settore affrontando un problema nuovo e più impegnativo: "Possiamo trasformare un'immagine di input arbitraria e renderla più memorabile?". Per formulare correttamente questa domanda si richiede l'esistenza di misure di memorabilità. I metodi per aumentare automaticamente la memorabilità dell'immagine avrebbero un impatto in molti campi di applicazione, come l'istruzione, i giochi o la pubblicità. Per affrontare il problema, introduciamo un approccio ispirato al paradigma dell'editing-by-applying-filters, adottato in applicazioni di fotoritocco come Instagram e Prisma. Gli utenti delle due app devono generalmente passare in rassegna i filtri disponibili prima di trovare la soluzione desiderata e si tratta di un processo che trasforma l’editing in un'attività che richiede risorse e tempo. Nel lavoro svolto ai fini di questa tesi, invertiamo il processo: data un'immagine in ingresso, ci proponiamo di recuperare automaticamente un insieme di “style seed”, cioè un insieme di immagini di stile che, applicate all'immagine in ingresso attraverso un algoritmo di neural style transfer, fornisce il massimo aumento della memorabilità. Di conseguenza, dimostriamo che è possibile recuperare automaticamente i migliori style seed per una determinata immagine, riducendo così notevolmente il numero di tentativi umani necessari per trovare una buona corrispondenza. Inoltre, dimostriamo l'efficacia dell'approccio proposto con esperimenti sul dataset LaMem, disponibile al pubblico, eseguendo sia una valutazione quantitativa che uno studio sugli utenti. Per dimostrare la flessibilità del framework proposto, analizziamo anche l'impatto delle diverse scelte di implementazione, come l'utilizzo di diversi metodi state-of-the-art di neural style transfer. Infine, mostriamo diversi risultati qualitativi per fornire ulteriori approfondimenti sul legame tra stile dell'immagine e memorabilità. Questo approccio nasce dai recenti progressi nel campo della sintesi delle immagini e adotta una deep architecture per generare un'immagine memorabile da una data immagine di input e da uno style seed. È importante sottolineare che per selezionare automaticamente lo stile migliore, basandosi anche su modelli deep, viene proposta una nuova soluzione learning-based. La valutazione sperimentale, condotta su benchmark pubblicamente disponibili, dimostra l'efficacia dell'approccio proposto per la generazione di immagini memorabili attraverso la selezione automatica degli style seed.

Bildgruppen på enheten för dokument och informationsteknik hos SKL har ett behov av att kunna förbättra bilder med extremt låg upplösning. Detta bildmaterial kan komma från diverse övervakningskameror där det intressanta objektet endast utgör en väldigt liten del, i detta fall registreringsskyltar på förbipasserande bilar. Att skapa en högupplöst bild av en registreringsskylt utav ett fåtal lågupplösta bilder är ett välkänt problem med ett flertal förslag på metoder och lösningar. I denna rapport kommer vi att undersöka vilka metoder som passar bäst vid bildförbättring av registreringsskyltar. Vi kommer även att skapa ett användargränssnitt där man kan läsa in en bild och välja mellan att automatiskt hitta registreringsskylten i bilden eller att manuellt klippa ut den. Efter att man erhållit en bild innehållandes endast registreringsskylten ska de olika implementerade bildförbättringsmetoderna kunna användas. Slutligen diskuteras vilka för- och nackdelar de respektive metoderna har. Förslag på eventuella förbättringar och hur man kan utveckla dessa metoder vidare presenteras därtill.

<p>Technologies in image acquisition have developed and often provide image volumes in more than two dimensions. Computer tomography and magnet resonance imaging provide image volumes in three spatial dimensions. The image enhancement methods have developed as well and in this thesis work 3D image enhancement with filter networks is evaluated.</p><p>The aims of this work are; to find a method which makes the initial parameter settings in the 3D image enhancement processing easier, to compare 2D and 3D processed image volumes visualized with different visualization techniques and to give an illustration of the benefits with 3D image enhancement processing visualized using these techniques.</p><p>The results of this work are;</p><p>1. a parameter setting tool that makes the initial parameter setting much easier and</p><p>2. an evaluation of 3D image enhancement with filter networks that shows a significant enhanced image quality in 3D processed image volumes with a high noise level compared to the 2D processed volumes. These results are shown in slices, MIP and volume rendering. The differences are even more pronounced if the volume is presented in a different projection than the volume is 2D processed in.</p>