<p>För att kunna rekonstruera en tredimensionell volym av en hjärna avbildad med hjälp av magnetresonanstomografi (MRI) behöver man korrigera varje snittbild i förhållande till varandra, beroende på oundvikliga rörelser hos den röntgade patienten. Detta förfarande kallas bildregistrering och idag använder man sig primärt av en metod där en bild utses till referensbild och därefter anpassas närliggande bilder, som antas ha en minimal avvikelse, till referensen.</p><p>Syftet med detta examensarbete är att använda en annan metod vanligen utnyttjad inom datorseende för att estimera ett rörelsefält utifrån en vanlig videosekvens, genom att följa markörer som indikerar rörelse. Målet är att skapa en robust estimering av huvudets rörelse, som då kan användas för att skapa en mer noggrann korrigering och därmed också en bättre rekonstruktion.</p> / <p>Before reconstructing a three dimensional volume from an MR brain imaging sequence there is a need for aligning each slice, due to unavoidable movement of the patient during the scanning. This procedure is known as image registration and the method used primarily today is based on a selected slice being the reference slice and then registrating the neighbouring slices, which are assumed to be of minimal deviation.</p><p>The purpose of this thesis is to use another method commonly used in computer vision - to estimate the motion from a regular videosequence, by tracking markers indicating movement. The aim is to create a robust estimation of the movement of the head, which in turn can be used to create a more accurate alignment and volume.</p>

<p>This thesis investigates how disparity estimation may be used to visualize an object on a 3D-screen. The first part looks into different methods of disparity estimation, and the second part examines different ways to visualize an object from one or several stereo pairs and a disparity map. Input to the system is one or several stereo pairs, and output is a sequence of images of the input scene but from more angles. This sequence of images can be shown on Setred AB's 3D-screen. The system has high real time demands and the goal is to do the disparity estimation and visualization in real time.</p><p>In the first part of the thesis, three different ways to calculate disparity maps are implemented and compared. The three methods are correlation-based, local structure-based and phase-based techniques. The correlation-based methods cannot satisfy the real-time demands due to the large number of 2D-convolutions required per pixel. The local structure-based methods have too much noise and cannot satisfy the quality requirements. Therefore, the best method by far is the phase-based method. This method has been implemented in Matlab and C and comparisons between the different implementations are presented.</p><p>The quality of the disparity maps is satisfying, but the real-time demands cannot yet be fulfilled. The future work is therefore to optimize the C code and move some functions to a GPU, because a GPU can perform calculations in parallel with the CPU. Another reason is that many of the calculations are related to resizing and warping, which are well-suited to implementation on a GPU.</p>

<p>The possibility to use real-time computer vision in video sequences gives many opportunities for a system to interact with the environment. Possible ways for interaction are e.g. augmented reality like in the MATRIS project where the purpose is to add new objects into the video sequence, or surveillance where the purpose is to find abnormal events.</p><p>The increase of the speed of computers the last years has simplified this process and it is now possible to use at least some of the more advanced computer vision algorithms that are available. The computational speed of computers is however still a problem, for an efficient real-time system efficient code and methods are necessary. This thesis deals with both problems, one part is about efficient implementations using single instruction multiple data (SIMD) instructions and one part is about robust tracking.</p><p>An efficient real-time system requires efficient implementations of the used computer vision methods. Efficient implementations requires knowledge about the CPU and the possibilities given. In this thesis, one method called SIMD is explained. SIMD is useful when the same operation is applied to multiple data which usually is the case in computer vision, the same operation is executed on each pixel.</p><p>Following the position of a feature or object in a video sequence is called tracking. Tracking can be used for a number of applications. The application in this thesis is to use tracking for pose estimation. One way to do tracking is to cut out a small region around the feature, creating a patch and find the position on this patch in the other frames. To find the position, a measure of the difference between the patch and the image in a given position is used. This thesis thoroughly investigates the sum of absolute difference (SAD) error measure. The investigation involves different ways to improve the robustness and to decrease the average error. One method to estimate the average error, the covariance of the position error is proposed. An estimate of the average error is needed when different measurements are combined.</p><p>Finally, a system for camera pose estimation is presented. The computer vision part of this system is based on the result in this thesis. This presentation contains also a discussion about the result of this system.</p> / Report code: LIU-TEK-LIC-2007:5. The report code in the thesis is incorrect.

<p>Initially an analytical closed-form inverse kinematics solution for a 5 DOF robotic arm was developed and implemented. This analytical solution proved not to meet the accuracy required for the shape sorting puzzle setup used in the COSPAL (COgnitiveSystems using Perception-Action Learning) project [2]. The correctness of the analytic model could be confirmed through a simulated ideal robot and the source of the problem was deemed to be nonlinearities introduced by weak servos unable to compensate for the effect of gravity. Instead of developing a new analytical model that took the effect of gravity into account, which would be erroneous when the characteristics of the robotic arm changed, e.g. when picking up a heavy object, a learning approach was selected.</p><p>As learning method Locally Weighted Projection Regression (LWPR) [27] is used. It is an incremental supervised learning method and it is considered a state-ofthe-art method for function approximation in high dimensional spaces. LWPR is further combined with visual servoing. This allows for an improvement in accuracy by the use of visual feedback and the problems introduced by the weak servos can be solved. By combining the trained LWPR model with visual servoing, a high level of accuracy is reached, which is sufficient for the shape sorting puzzle setup used in COSPAL.</p>

<p>Coated solid bleached board belongs to the top-segment of paperboards. One important property of paperboard is the printability. In this diploma work a specific print defect, oriented mottle, has been studied in association with Iggesund Paperboard. The objectives of the work were to develop a method for analysis of the dark and light areas of oriented mottle, to analyse these areas, and to clarify the effect from the print, coating and paperboard surface related factors. This would clarify the origin of oriented mottle and predict oriented mottle on unprinted paperboard. The objectives were fulfilled by analysing the areas between the dark halftone dots, the amount of coating and the ink penetration, the micro roughness and the topography. The analysis of the areas between the dark halftone dots was performed on several samples and the results were compared regarding different properties. The other methods were only applied on a limited selection of samples. The results from the study showed that the intensity differences between the dark halftone dots were enhanced in the dark areas, the coating amount was lower in the dark areas and the ink did not penetrate into the paperboard. The other results showed that areas with high transmission corresponded to dark areas, smoother micro roughness, lower coating amount and high topography. A combination of the information from these properties might be used to predict oriented mottle. The oriented mottle is probably an optical phenomenon in half tone prints, and originates from variations in the coating and other paperboard properties.</p>

<p>Teching aid in form of schoolbooks plays an importent role. The teacher uses schoolbooks as a tool in its tutoring. I have examined how and whether the pictures in the schoolbooks are presenting people and ethnic minorities biased and stereotypical. The material I have examined consisted of a series of schoolbooks (So-direkt 1-3, samhällskunskap av Bonnier 2003).</p><p>To find out if that is the case I have used a theory by Stuart Hall, his theory is about people and how they are represented in different types of media, like pictures and text. I studied it by using the method of image analysis where I investigated and interpreted the denotation and konnontation of the pictures. Through image analysis I have looked at, examined and interpreted pictures that shows people in rich or poor contexts. From that I discoverd that people does not depict stereotypicaly in the pictures, but i did find myself seeing that people are positivly and/or negatively depicted depending on who are shown in the picture. Ethnic minorities are underrepresented in this series of books. Pictures which are depicting ethnic minorities are only shown in a total of nine pictures. In the way that these books are presenting Swedes they are only showing one typ of Swede. That Swede is nearly always light-skinned and has either blond or brown hair. The diversity that exist in reality is not representet in these schoolbooks. These schoolbooks tend to present certain people more positively then others through the pictures they depict. When it comes to show the environment and people it is done by showing the swedish population living in wealth. Swedish people are seen consuming and producing commodities, you can also see a variety of technological machines as tv:s, cellphones, phones, cars, computers, airplanes and trains surrounding them. Other people who are dark-skinned tend to be presented in a poor enviroment and in a negativ way in pictures.</p>

<p>This report contains a description on how it is possible to guide a robot along an edge, by using a camera mounted on the robot. If stereo matching is used to calculate 3Dcoordinates of an object or an edge, it requires two images from different known positions and orientations to calculate where it is. In the image analysis in this project, the Canny edge filter has been used. The result from the filter is not useful directly, because it finds too many edges and it misses some pixels. The Canny edge result must be sorted and finally filled up before the final calculations can be started. This additional work with the image decreases unfortunately the accuracy in the calculations. The accuracy is estimated through comparison between measured coordinates of the edge using a coordinate measuring machine and the calculated coordinates. There is a deviation of up to three mm in the calculated edge. The camera calibration has been described in earlier thesis so it is not mentioned in this report, although it is a prerequisite of this project.</p>

<p>For marine sciences sometimes there is a need to perform underwater photography. Optical properties of light cause severe quality problems for underwater photography. Light of different energies is absorbed at highly different rates under water causing significant bluishness of the images. If the colour dependent attenuation under water can be properly estimated it should be possible to use computerised image processing to colour correct digital images using Beer’s Law.</p><p>In this thesis we have developed such estimation and correction methods that have become progressively more complicated and more accurate giving successively better correction results. A process of estimation of downwelling attenuation coefficients from multi or hyper spectral data is a basis for automatic colour restoration of underwater taken images. The results indicate that for each diving site the unique and precise coefficients can be obtained.</p><p>All standard digital cameras have built in white balancing and colour enhancement functions designed to make the images as aesthetically pleasing as possible. These functions can in most cameras not be switched off and the algorithms used are proprietary and undocumented. However, these enhancement functions can be estimated. Applying their reverse creates un-enhanced images and we show that our algorithms for underwater colour correction works significantly better when applied to such images.</p><p>Finally, we have developed a method that uses point spectra from the spectrometer together with RGB colour images from a camera to generate pseudo-hyper-spectral images. Each of these can then be colour corrected. Finally, the images can be weighted together in the proportions needed to create new correct RGB images. This method is somewhat computationally demanding but gives very encouraging results.</p><p>The algorithms and applications presented in this thesis show that automatic colour correction of underwater images can increase the credibility of data taken underwater for marine scientific purposes.</p>

Image segmentation continues to be a fundamental problem in computer vision and image understanding. In this thesis, we present a Bayesian network that we use for object boundary detection in which the MPE (most probable explanation) before any evidence can produce multiple non-overlapping, non-self-intersecting closed contours and the MPE with evidence where one or more connected boundary points are provided produces a single non-self-intersecting, closed contour that accurately defines an object's boundary. We also present a near-linear-time algorithm that determines the MPE by computing the minimum-path spanning tree of a weighted, planar graph and finding the excluded edge (i.e., an edge not in the spanning tree) that forms the most probable loop. This efficient algorithm allows for real-time feedback in an interactive environment in which every mouse movement produces a recomputation of the MPE based on the new evidence (i.e., the new cursor position) and displays the corresponding closed loop. We call this interface "object highlighting" since the boundary of various objects and sub-objects appear and disappear as the mouse cursor moves around within an image. / Graduation date: 2004

For marine sciences sometimes there is a need to perform underwater photography. Optical properties of light cause severe quality problems for underwater photography. Light of different energies is absorbed at highly different rates under water causing significant bluishness of the images. If the colour dependent attenuation under water can be properly estimated it should be possible to use computerised image processing to colour correct digital images using Beer’s Law. In this thesis we have developed such estimation and correction methods that have become progressively more complicated and more accurate giving successively better correction results. A process of estimation of downwelling attenuation coefficients from multi or hyper spectral data is a basis for automatic colour restoration of underwater taken images. The results indicate that for each diving site the unique and precise coefficients can be obtained. All standard digital cameras have built in white balancing and colour enhancement functions designed to make the images as aesthetically pleasing as possible. These functions can in most cameras not be switched off and the algorithms used are proprietary and undocumented. However, these enhancement functions can be estimated. Applying their reverse creates un-enhanced images and we show that our algorithms for underwater colour correction works significantly better when applied to such images. Finally, we have developed a method that uses point spectra from the spectrometer together with RGB colour images from a camera to generate pseudo-hyper-spectral images. Each of these can then be colour corrected. Finally, the images can be weighted together in the proportions needed to create new correct RGB images. This method is somewhat computationally demanding but gives very encouraging results. The algorithms and applications presented in this thesis show that automatic colour correction of underwater images can increase the credibility of data taken underwater for marine scientific purposes.