Media Technology 2010


Title: Spatio-Temporal Colour Correction of Strongly Degraded Films  

The archives of motion pictures represent an important part of precious cultural heritage. But, these collections of cinematography are vulnerable to different types of distortions—specially, the effect caused by chemical support on which they are recorded becomes unstable with time, unless they are stored at low temperatures. Some defects on color movies, such as bleaching, are irreversible and hence it is beyond the capability of digital restoration process. We propose here an automatic color correction technique which eventually automates the color fading restoration process. The proposed method is based on STRESS model, an automatic image enhancement technique that deals with correcting color images which are more or less perceptually acceptable to the Human Visual System. We also propose some preprocessing techniques which will be applied to the distorted images in prior to apply STRESS algorithm. These preprocessing techniques, which include Principle Component Analysis (PCA) and some sort of saturation enhancement, will ultimately make the resulting image more appealing and acceptable to Human Visual System.


Digital film restoration, automatic colour correction, colour constancy, image enhancement, spatial color algorithms.

Problem description:  

For the last few decades, the cinematographic archives have been considered as the principal medium for the documentation of research and collections. But more and more old photographs become collector's items themselves. Their unique importance for research relating to the history of civilization is due to the fact that they are the only authentic sources for certain subjects or periods of time. However, in most case the storage conditions of the pictures are unsuitable, and a lot of them are in a poor state, because photographic film and paper are unstable media.

Color photographs are most vulnerable to the degradation process. The dyes of common chromogenic processes are chemically rather unstable. Photographic films contain not only dyes but also a few other components (sensitizers, color couplers, stabilizers etc.) which can alter with time. Since the 1950s, color film became the standard on which millions of cinematographic works were recorded. A couple of decades later, it turned out that this process was chemically unstable, causing the fading of whole film stocks with time. Usually, a bleached color release print is the only available record of a film. Since the bleaching phenomenon is irreversible, photochemical restoration of faded prints is not possible, hence the incontestability of digital color restoration.

We propose here a technique for the digital color restoration of faded movies that is based on a perceptual approach, inspired by some adaptation mechanisms of the human visual system (HVS). Particularly, we will be dealing with a combination of some preprocessing techniques for distorted images which will be further processed by the image enhancement model called STRESS. The final processed image should have more appealing result from the Human Visual System’s point of view.


Title: Scene recognition for improved aesthetic quality inference of photographic images  

The novel field of aesthetic quality inferencing of natural images deals with the automatic assessment of the aesthetic value of a given photograph, by either numerically rating it, or by classifying it as a professional, high-level picture or as a low quality snapshot. Based on the extraction of low-level features, a number of authors have tried to bridge the aesthetic gap given by the inherently subjective nature of aesthetics and, using machine learning techniques, set the basis for the development of a field with potential applications in areas as diverse as CBIR, management and editorial work or consumer photography. Their methods range from the opaque, black-box approach to more content-aware procedures, which define their feature set after well-established photographic techniques and build their success upon the prior identification of the photographic subject.

Our approach aims to go one step further in the understanding of the image content and, under the assumption that different subject categories require different composition techniques, introduces an additional scene type classification step, which, combined with the use of state of the art feature sets, should yield a significant improvement over the current performance results.

Keywords: Aesthetics, image quality, aesthetic quality, photography, image classification, pattern recognition, image processing, imaging


Title: Towards a perceptual metric for video quality assessment  


Tittle: Calenders as user context providers in a learning environment  

Presentation (pdf) 

User context is considered a key component when adapting a learning material to the users preferences. Users are different and have different learning styles when exposed to learning material, the issue concerning the shifting of learning environment from being an arena existing on the systemÕs premisses, to be an arena that adjusts on the behalf of the users, is quite extensive. But guiding each user during a learning process is an important aspect of adapting the learning material, and a crucial part of making the learning environment user-oriented. User calendars could be a valuable resource for extracting a users context, but we do not know how reliable they are, or if the information they contain are sufficient. In other words, can calendars be considered a source that contribute to defining a users context? Ultimately we do not know if a context-aware LMS that utilize calendars will aid the process of moving the learning environment from the systems preferences to the users. Providing learning objects that correspond to the users preferences could be an important step in this area of research.

The aim of this thesis is to provide knowledge that contribute to this work, by locating the information variables about the user context and providing this information to the learning management system. Several challenges are connected to this; firstly, to find a set of variables that can be used to state information about the user context, and finding the sources where these variables can be extracted from.

The thesis scope includes a study on user calendars provided by a test group, and the design of a simple prototype that provides learning material according to the user context.


Title: Smoothness of color transforms  

Presentation (pdf) 

Color image quality is an important factor in various media such as digital cameras, displays and printing systems. The employment of different color imaging media leads to a constant problem that color reproduction from each medium produces color differently. It makes all different manufactories to focus on the technology to achieve successful cross-media reproduction. In this case the image reproduction quality depends on processes of device characterization. The device characterization and profiling are central processes which allow predicting a result of device reproduction according to the known input and provide communication of devices. The Look-up tables (LUTs) are the most common empirical approach for device characterization, and are the basis for ICC profiles. Smooth LUTs-based color conversion in device characterization is an important factor for achieving high quality of the reproduced color image. Such factors as LUTs size, interpolation methods and unavoidable noise in color measurement process and unstable printing process influence on smoothness of LUTs-based color transforms, and result in appearing artifacts on final reproduced images.

The main goal of this project is to find a way of quantifying smoothness of color transforms through analysis of LUTs based device characterization process and the factors which affect on it and test different image quality metrics. It requires conducting a number of experiments for determining and estimating thresholds of factors for predicting unsmooth transform and corresponding human perceptual evaluation of smoothness of color transforms in comparison with quantitative evaluation results provided by different image quality metrics. The evaluation smoothness of LUTs-based color transforms will allow to avoid undesirable results in image color reproduction systems such as artifacts and distortions of image content and improving device characterization process for achieving smooth color transform. It is an attempt to continue work in direction of getting qualitative reproduction of color images fast and economically and developing technology advances for obtaining it.

Keywords:  device characterization, calibration, color conversion, LUT, look up table, image color quality, smoothness, ICC profiles, color management


Title: Using augmented reality and tangible user interfaces in a primary school learning situation  

Presentation (pdf) 

Augmented Reality (AR) is a technology where synthetic objects can be placed into the real world, in real time. This means that 3D objects can be manipulated directly without using the traditional mouse and keyboard paradigm. This master thesis will look at using these approaches in a learning situation. There might be several advantages in doing so. Pupils at the primary school do not have the same reasoning and experience adults have, so certain subjects might be difficult to understand without seeing it directly. AR is a good technology for visualizing things, in addition by making the interface tangible a learning system might encourage collaboration among the pupils. It is also well known that learning by doing tends to increase learning. In the thesis some sample applications will be developed and tested on a group of 6th graders.

Another problem in AR systems is that the knowledge needed for creating content is quite high. Looking at involving teachers in the content creation process will also be looked into.


Title: Analysis of geographical information systems(GIS) in geography learning.  

Presentation (pdf) 

Norwegian high schools have recent years gone through new reforms which all have stated that the use of new technology in education is important. Based on this, and the fact that the use of geographical information systems (GIS) has rapidly increased recent years, is this master thesis aimed at the use of GIS in education. The work will mainly consist of the development of a GIS application suited to a specific part of the curriculum in the geography course held in Norwegian high schools. The application will then be evaluated against the ordinary resources available for the students, and evaluated by teachers responsible for the course. Hopefully will this thesis end in a greater knowledge about how GIS can be used in education, and also say something about the effect of using a GIS application as a replacement, or an additional tool in education.


Title: Utilizing students’ input to create and manage learning object metadata in educational system  

Presentation (pdf) 


Learning Objects have been used in the educational practices for many years. However, the issues of the creation and the reusing of learning objects are still discussed in the current stage. Most of these questions are related to the learning object metadata. The generation of learning object metadata is great crucial for the reusing of learning objects. Manual and automatic metadata generation approaches are two major methods. Folksonomy is one kind of manual methods. Although manual metadata generation method is time-consuming, it still becomes popular because more and more social software applications use their users’ defined classification to tag learning object. In educational system, students’ inputs can be regarded as one set of folksonomy data which would be used to display learning object with more descriptions. Therefore, how to use students’ inputs to generate effective metadata for learning object is concerned in the project.

Learning Object Management System can generate learning object metadata according to the creation of learning objects and students’ input for each. For the same learning object, there are some different understandings between teacher and students. Students’ inputs contain tags and the evaluations for the learning object, their comments and advices are looked as this part as well. Thus, the main research question in this thesis is whether we can regard students’ inputs—folksonomy data—as high-quality metadata for learning objects in educational system. Moreover, whether the reusing efficiency of learning objects is improved by students’ inputs would be analyzed in the project as well.

The object of this project is to analyse the students’ input in order to improve the reusing of learning objects, and then test whether this set of metadata is reliable or not in the educational system. The research questions focus on three parts, the first one about automatic generation based on ontology which is generated by the teachers, the second one concerning folksonomy metadata which is created by students, and the last but not least one regarding analysis the difference of the learning object metadata generated in the pre-mentioned two part information.


Title: Salient color feature extraction for image retrieval  

Presentation (pdf) 

Color is an important visual cue and includes many key features in the image content analysis. Most current color descriptors such as the dominant color descriptor, still have limitations to depict how human color perception looks at images. One big issue is that, the local spatial organization of the visual scene, which contains the prominent colors, are not the objects the user intends to find. Much redundancy has been incurred in the image descriptors leading to the great complexity in the similarity measure. Therefore, in this thesis we propose a compact global color descriptor, which is able to extract the distinct features in the visual scenery. The objective falls into two intertwined parts: Firstly, a salient feature detection approach is proposed based on the opponent color space. All the dominant colors extracted by the Particle Swarm Optimization (PSO), are transformed to an iso-salient space to boost their color saliencies. An evaluation test is performed to compare the extracted salient regions with the human gaze map. Secondly, a novel spatial descriptor is used to rescale the local regions according to the importance in the salient feature map. Finally, a combined color descriptor is evaluated in a large database to show its strength.

Keywords:  Content-based indexing and retrieval, visual saliency, color imaging, perceptual color descriptor


Title: Using a geographical map data to improve navigation for visually impaired and blind pedestrians  

Presentation (pdf) 


The concept of Universal Design was first introduced in the United States by architect Ronald L. Mace. He explained the Universal Design in an article in Designers West, this was in 1985 and is the first documented use of the term. Where Mace described the concept of universal design as "designing all products and the built environment to be aesthetic and usable to the greatest extent possible by everyone, regardless of their age, ability, or status in life". People who can see uses their sight to orient themselves and gain information about their surroundings. Blind or visually impaired people have little or none of this information, which hampers their ability to get information about their surroundings and to navigate within them. A common task like going to the store becomes a difficult process since they can not see and orientate. Alternative methods are used, such as listening and touch, as a substitute for the missing sense.

The central idea for this thesis is to gather information and experience about the processes of creating a navigation-system for blind or visually impaired people. A system will be developed, which will be based on a mobile phone equipped with GPS, and guide the user by sending audio feedback such as speech and sound. The system will be developed by using the Java programming language and the Android platform for mobile phones, combined with a database to store geographical information.

This raises a couple of research question, For example what effect this system will have on people who are blind? Also how does the system work compared to typically guidance such as a trained dog or by using a blind-cane? Results will be obtained by testing the system on people who are visually impaired and by using questionnaires.


Title: TouristGuide an iPhone application to make tourism around Barcelona


Human beings, by nature, need to move, find new frontiers. In 2009, Spain had 52.2 million tourists that left 48.242 million euros. They normally go through the city with a map, which has been previously added the Points of Interest. However the emergence of mobile devices with built-in GPS has opened another option to visit a city.
In this direction appears TouristGuide, a free application for iPhone. This application has been programmed with Objective-C language and has the ability to generate routes through a series of preloaded sights. TouristGuide wants the tourist to feel comfortable visiting Barcelona, avoiding the feeling of being lost, thus with this application will always know where he is and the route to follow.


Title: Education through videoconference – User friendly and effective videoconference teaching  

Videoconference has made the world smaller. It is common now to use videoconference in the educational system too. Universities and institutions have invested in videoconference equipment mainly to use for distance education, but also for meetings with other institutions that they cooperate with. Does this investment really pay for itself? Is it used as much as it should? Is a videoconference lecture as effective as a traditional classroom lecture? The technology is evolving from day to day, but there are still some obstacles until a lecture through videoconference can be realized with the same confidence as a traditional lecture.

There is not easy to be interactive and have full concentration during a VC-class. Teachers do not see all the students, or more specifically their body language and their reactions. Students on their side feel that this is just a presentation from someone far away, it becomes boring after the first lecture hour.

The aim of this project is to design a prototype for effective videoconference within education. The purpose is to make education through this type of medium more user-friendly and pedagogically more efficient. The motivation for this master thesis is to provide new knowledge and better solutions for lecturing through distance, mainly through videoconference, so students can benefit more from this kind of lectures.


Title: Kubelka Munk Theory for Time Saving Spectral Printer Modelling  

Presentation (pdf) 

Abstract :  

Our work particularly focus’s on the quality of spectral colour image reproduction of the printing world, more specifically on the spectral printer modelling. There has been different spectral printer models developed so far. The spectral Neugebauer model is the spectral printer model which is the generalizationof the Murray–Davies equations to include more thantwo reflectances. In this model the reflectance of the output from the printer can be computed as a convex combination of Neugebauer Primaries.

To get these reflectances of Neugebauer Primaries one has to find out the way to print out them so that he/she could measure the reflectances. It has the following problems. The number of patches we are going to print increases exponentially with the number of colorants of the printer since the number of possible combination of colorants is given by 2n , where n  is the number of colorants of the printer. For Spectral printer with more than 7 or 8 colorants, printing all those Neugebauer Primaries and measure them will consume lots of time and materials. The other problem is that finding out the way to really print out all possible combinations is very difficult. There are some charts which incorporate all Neugebauer Primaries of CMYK printer. The problem will begin when we consider printers with more colorants than these 4. There are no charts which actually incorporate Neugebauer Primaries of 8 channel printer. In this case the researcher forced to have the SDK of the printer and find out some way to order the printer to print anything he/she wants.

In this thesis we tested Kubelka-Monk theory in order to estimate those Neugebauer Primaries so that we could save our resources, power and time. We used different kinds of papers for test and for recommending more cheapest and accurate way of using this theory. Then we spectrally model the CMYK HP Deskjet 1220C printer and Xerox phaser 7760 laser printer using spectral Neugebauer model and Yule-Nielsen modified spectral Neugebauer (YNSN) model. Using these models we show that how much more improvements we could actually get by estimating Neugebauer Primaries using Kubelka-monk theory rather than using real measurements of the Neugebauer Primaries.


Title: GML validation based on Norwegian standard  

Presentation (pdf) 


One UML model in geography field can be realized in many formats. For more than two decades, SOSI is the standard format for realizing the UML model in Norway. But as the technologies develop, GML is regarded as the standard and most used geographic data format worldwide. Therefore, to be able to share geographic data and be compatible with the world, the UML model based on Norwegian standard should also be realized in GML format. One way of doing this is to make the GML files from the very beginning according to the models, which is very costly; the other way is transforming the already made SOSI files to GML files, which is time and effort saving, however with some problems. On one hand, there is not only one software doing the transformation between SOSI format and GML format, thus the GML files are not expressed in the same structure and elements. On the other hand, it is unsure that the GML files are consistent with the UML models. In one word, the GML file field is messy that whether one GML file is qualified is unknown.

The aim of this project is to build up a GML validation mechanism based on Norwegian standard so that the GML file field in Norway can be under control. The validation is between a GML file and its schema, the schema is an XML file gained from the graphic UML model. So by this validation, a GML file’s quality is not longer under unknown status. Therefore, the GML file field is in order. Furthermore, it will be more efficient to share the geographic data with other countries with qualified GML files.


Title:   A comparison of list and scatterplot interfaces for data comparison on mobile devices  

Presentation (pdf) 


Viewing and navigating large visual information spaces on mobile phones is difficult due to their small screens and cumbersome input. This will hold true for the foreseeable future, as one of the main aspects of mobile phones is to retain a small physical size for the sake of portability.
While much research has gone into how content can be visualized and navigated on small screens, there doesn't seem to be much information on how the on-screen contents can be compared to each other.
This thesis focuses on how we can compare the data that is being browsed, and how the relation between content with similar properties can be visualized to allow quick identification and selection of the items that best match the values we are looking for.


Title:   Conceiving a fast and practical multispectral-stereo system  

Presentation (pdf) 

Multispectral and 3D imaging are two complimentary imaging technologies with many advantages and great potential for their widespread use in the future, if we can make them faster and more practical. This thesis aims at conceiving such a fast and practical two-in-one multispectral-stereo system.

Multispectral imaging systems remedy the problems of conventional three channel (RGB) color imaging like metamerism and dependency on acquired conditions and at the same time presents high spatial and spectral resolution. A multispectral image is composed of several monochannel images of the same object, each image holds data about a specific wavelength depending on the filter used. The major problems of existing multispectral imaging systems are that they are slow as they require multiple takes and/or they are quite expensive, contributing to the current lack of widespread use in the consumer segment. This thesis will explore creating a fast, practicable and affordable multispectral system with the use of two commercially available digital cameras. Each camera is equipped with an optical filter. These two filters are chosen so that they modify and spread the sensitivities of the cameras so that they become well spaced throughout the visible spectrum covering complementary wave-bands, thus giving rise to six channel multispectral system.

Like multispectral imaging, 3D imaging systems are also gaining popularity and are of great use in imaging fields. It would be great advantage if we can have an integrated system capable of both multispectral and at the same time 3D imaging. This thesis, therefore, aims at this in conceiving such a multispectral-stereo system. The two cameras modified with appropriate filters that forms six channel multispectral system are used in stereoscopic configurations to acquire depth information making it capable of 3D imaging as well. This leads to a faster and practicable and at the same time affordable multispectral-stereo systems.

Such a system could be used for many applications, for example for 3D artwork object acquisition. Knowing the spectral reflectance allows us to simulate the appearance of a 3D object under any virtual illuminant. Moreover, it lets us store this valuable information for future restoration.

Keywords:  Multispectral, stereoscopic, Imaging,3D, RGB camera, filter selection


Title: Context-aware recommender systems  

Presentation (pdf) 


Recommender systems are systems that provide recommendations to a user based on information gathered about that user or by finding other similar users. Only a few years ago, these recommender systems were solely dependent on either explicit information given by the user or implicitly gathered information such as user patterns- and actions. During these last few years we have seen a slightly increased interest incorporating context within recommender systems in order to provide better recommendations. There has been however, little contribution to semantically enabled contextual recommender systems based on ontologies. The same applies to what kind of contextual features that are relevant to such a recommender system in different e-commerce settings. By exploiting more of the information that lies in the users' context, we can use semantic web technologies to infer more about what the user wants. For instance, if a person at work is visiting an e-store, the person might want to only see recommendations related to work and personal-related recommendations at home. Applying contextual features into recommender systems could convert window-shoppers into buyers, increase cross-sell and customer satisfaction and ultimately generate more revenues to the company.

During this master thesis project it is hoped to find some of the contextual features that are actually relevant to use in a given e-commerce setting. In addition, a semantically context-aware recommender system based on ontologies will be developed and measured quantitatively to analyze its performance compared to a non-contextual one, and see if there are any improvements.


Title: Probabilistic framework for multi-target tracking using multi-camera: applied to fall detection  

Presentation (pdf) 


The aim of the project is to detect fallingpeople in the surveillance videos in nursing home and raise an alarmto notify the concerned person about the event.The task is to develop a system that will monitor people trackingusing multiple cameras. Some work had already been done at university:single camera person tracking. So, need to improve the existing singletracking algorithm and propose a modified version for multiple cameratracking. The fusion of the information from multiple cameras andtheir analysis for fall detection is the main task of this project.


Title: Using WFS-T in combination with geolocation for online transactional editing of geographic features through mobile browsers  

Presentation (pdf) 


While online maps have been widely used in the society, the number of providers who develop applications to serve those users needs have also been increasing. To make use of data or exchange data from each provider, some standards are involved. The OpenGIS Consortium (OGC) is an association offering certain standards for geoprocessing which can be helped in developing and implementing software.

The purpose of this project is to use WFS-T in combination with geolocation for online transactional editing of geographic features through mobile browsers on mobile phones based on one of OGC standards called WFS (Web Feature Service), which WFS-T (Transactional Web Feature Service) is a part of WFS. With WFS-T, users can perform transactional operations such as create, update and delete geographic features. Geographic features can be described as a set of properties which may include a name, type and value. For example, one country can be defined as a geographic feature containing the following properties: {country name, country capital, country lonlat, country population}; each property can have different type and value such as country name has a type as string and has a value as Norway while country population has a type as number and has a value as 4882600.

In this implementation, GeoServer is selected to be used as GIS server software, PostgreSQL with PostGIS is used as datastore and OpenLayers is used in creating web-based geographic application. Additionally, W3C Geolocation API is used to detect usersÕ locations from using their devices, in this case, an Android phone is used as a test device during the implementation. Once users launch their mobile browsers to connect to the Internet, the application will figure the usersÕ locations and show the position as a point on a map through their mobile browsers. Moreover, the amenity places around their area will be displayed. The prototype allows users to make online transactions such as insert a new point as a new place. When users finish and save their work, all data will be recorded in the database via WFS-T operations.

The project report will give readers some background information about geolocation, the OpenGIS Standards including concepts and technologies of GIS and mobile browsers, how the prototype is created, and also the results of the implementation to provide information for those who are interested in using WFS-T to edit geographic features through web browsers on mobile phones.