The rest of the paper is organizes as follows: In Section 2 we present the main notations used throughout the rest of the paper. In Section 3 we introduce the sequential and parallel non- overlapping domain decomposition method and show its convergence to a minimizer of the global problem. Details on the numerical implementation of the solvers for the proposed domain decomposi- tion methods are described in Section 4. Finally in Section 5 we show sequential and parallel numer- ical experiments for imagedenoising. We compare the computational performance of the proposed domain decomposition algorithm of the dual problem with the domain decomposition algorithm for the problem in (2) introduced in . Moreover, we show the successful application of the proposed algorithm for 3D medical data.
In this paper we present an alternative method for SAR imagedenoising, structure enhancement, and change detection based on the curvelet transform. Curvelets can be denoted as a two dimensional further development of the well-known wavelets. The original image is decomposed into linear ridge-like structures, that appear in different scales (longer or shorter structures), directions (orientation of the structure) and locations. The inﬂuence of these single components on the original image is weighted by the corresponding coefﬁcients. By means of these coefﬁcients one has direct access to the linear structures present in the image. To suppress noise in a given SAR image weak structures indicated by low coefﬁcients can be suppressed by setting the corresponding coefﬁcients to zero. To enhance structures only coefﬁcients in the scale of interest are preserved and all others are set to zero. Two same-sized images assumed even a change detection can be done in the curvelet coefﬁcient domain. The curvelet coefﬁcients of both images are differentiated and manipulated in order to enhance strong and to suppress small scale (pixel-wise) changes. After the inverse curvelet transform the resulting image contains only those structures, that have been chosen via the coefﬁcient manipulation. Our approach is applied to TerraSAR-X High Resolution Spotlight images of the city of Munich. The curvelet transform turns out to be a powerful tool for image enhancement in ﬁne-structured areas, whereas it fails in originally homogeneous areas like grassland. In the change detection context this method is very sensitive towards changes in structures instead of single pixel or large area changes. Therefore, for purely urban structures or construction sites this method provides excellent and robust results. While this approach runs without any interaction of an operator, the interpretation of the detected changes requires still much knowledge about the underlying objects.
Observe that (1.3) represents a so-called variational inequality of the second kind . Previously, in  a posterior residual based error estimates for a specific (and with respect to j a somewhat simpler) class of variational inequalities of the second kind were derived. In view of (1.3), our aim is to extend these estimates to situations where j models the TV-regularization term in imagedenoising and where the variational inequality is yet formulated in a di↵erent primal-dual fashion (see the subsequent section) which is in particular suitable for numerical solution schemes.
on two levels to the results using the normal multiscale analysis. The first approx- imations in the iterative procedure, i.e. the approximations with a large diffusivity are so far away from the data at the peak, that the multiresolution condition is not satisfied on the whole image initially, and later on the quarter of the image, where the peak is situated. Therefore the diffusivity a(x) is decreased in those regions, if the normal multiresolution criterion is applied. But note that it is sufficient to decrease a(x) at the peak, and a large diffusivity in the remaining pixels leads to an approximation which satisfies the multiresolution criterion on the whole par- tition P. Therefore the multiscale analysis on two levels yields smoother results and a better approximation of small features than the normal multiscale analysis. We have seen that the partition P S into dyadic squares is to be preferred to the
In the first chapter we describe some nonparametric regression methods and dis- cuss the problems concerning the selection of the smoothing parameters. In case of datasets with varying smoothness, estimators with a local smoothing parame- ter are preferred, naturally, to those with one global smoothing parameter. The smoothing parameter of the Nadaraya-Watson kernel estimator for instance can be localized. However it is not suitable for denoising two-dimensional datasets since it takes relatively long to compute it. Similar drawbacks of other known methods are pointed out in Chapter 1, to show that our method can be utilized with advantage. Indeed, not only the computing time is reduced considerably by the use of our method, but also smoother results can be obtained.
After spectral pre-processing, the Raman hyperspectral imaging data were subjected to a spatial pre-processing routine, edge-preserving denoising (EPD), which was recently proposed in the context of analysis of hyperspectral MALDI imaging mass spectrometry data . EPD is an operation in the image domain and plays a crucial role in the proposed spatial segmentation pipeline. The aim of EPD is to reduce noise-related pixel-to-pixel variation often unavoidable in Raman microspectroscopic imaging at the same time preserving small spatial features of the wavelength images. Note that the pixel-to-pixel variation is amplified when using the second derivative Raman spectra. For EPD, we used the locally-adaptive edge-preserving imagedenoising algorithm based on minimizing the total variation (TV) of an image . Informally speaking, TV of a gray-scale image is the sum of absolute differences of intensities at
3 Image edge model
When sound propagates through urban environments, there may be cases where the total sound ﬁeld cannot be fully described considering re ﬂections and diffractions sepa- rately. A homogeneous set of paths that only consists of re ﬂected or diffracted energy, even for very high orders, is not suf ﬁcient to generate a continuous sound ﬁeld. Algo- rithms determining homogeneous propagation paths may lead to fewer or even no audible paths, in contrast to hetero- geneous propagation paths, that allow for arbitrary combi- nations of re ﬂected and diffracted components. The implementation of a heterogeneous propagation path algo- rithm is signi ﬁcantly more challenging as it requires the integration of diffracting edges into an image model. There- fore, the Image Edge Model (IEM), depicted in Figure 1, is introduced. By mirroring edges along faces and applying the angle constraints for the diffraction coef ﬁcients (compa- rable to the re ﬂected angle of a specular reﬂection), com- bined re ﬂection and diffraction paths can be found. The classical Image Source Model (ISM) can be used to deter- mine re ﬂection paths as depicted in Figure 2a, which may lead to audible sound even if objects are occluding the line-of-sight. Homogeneous diffraction paths can be obtained by resolving the Equation System of Equal Angle (ESEA), which has been formulated by Tsingos et al.  in the context of a beam tracing approach. It states that the incoming and outgoing vectorial directions at interac- tion points I must maintain the same angle regarding the edge following:
dig nach außen? Die Glaubwürdigkeit hängt entschei dend davon ab, ob die Mitarbeiter selbst von der Auf gabe überzeugt sind und nach dieser «Vision» tatsächlich handeln. Die Kultur ist also nicht ein neuer Begriff für «Imagepflege», sondern die Gesamtheit der gelebten Werte und Handlungsprinzipien eines Unternehmens. Bei einem erfolgreichen Unternehmen spiegelt das von außen wahrgenommene Image die Unternehmenskultur wider, denn nur dann leistet das Unternehmen tatsäch lich das. was es verspricht.
Diese Überlegungen zusammenfassend kann man feststellen, daß sich die »Au- topoiesis« des Systems im Bereich der Werbung neben und mit der Unterscheidung Information/Nichtinformation über die Zweitcodierung Imagepositiv/Imagenegativ und einer darauf eingestellten Programmierung vollzieht. Mit diesen Operationen bestimmt sich der Systembereich selbst (die eigenen Objekte, die eigene Systemge- schichte) und seine Umwelt, also »Selbstreferenz« und »Fremdreferenz«. Obwohl die Werbung maßgeblich über ihre Zweitcodierung festgelegt ist, gilt es andererseits zu beachten, daß sie in ihren Inszenierungen immer auch mehr bzw. anderes als Image- Kommunikation betreibt. Wie andere Systeme reproduziert das System der Massen- medien (u.a. im Bereich der Werbung) keineswegs nur das »autopoietische Mini- mum« (Luhmann 1997, 406). Nicht alle Details einer Inszenierung müssen restlos auf Image bezogen sein. So kann die Werbung durchaus imageneutrale Informationen ]XP(LQVDW]EULQJHQXQGVLHPDFKWIDNWLVFKYRQGLHVHU0|JOLFKNHLWKlX¿J*HEUDXFK Daß ein Kühlschrank weniger Energie verbraucht als vergleichbare Produkte, daß eine politische Partei im Unterschied zur Konkurrenz (k)eine Erhöhung der Mehr- ZHUWVWHXHUDQVWUHEWRGHUGDNlXÀLFKH3URGXNWH]XEHVWLPPWHQKRKHQRGHUQLHGHUHQ Preisen zu erwerben sind, mögen Informationen sein, über die die Werbung berichtet. Als Werbung fungieren die verschiedensten Informationen jedoch nur, wenn sie auf der Basis visueller Kommunikationen als Attribute der beworbenen Objekte identi- ¿]LHUWZHUGHQXQG]XU4XDOL¿]LHUXQJGHUVHOEHQEHLWUDJHQ:LHHUZlKQWJHKWHVLKU inszenatorisch auch darum, Aufmerksamkeit herzustellen und die Erinnerung ihrer Mitteilungen wahrscheinlich zu machen. Und immer wieder (aber keineswegs prin- zipiell) muß sie die Glaubwürdigkeit ihrer Botschaften dramaturgisch berücksichti- gen. 189 In jedem Fall aber muß die Werbung auch der Herstellung von guten Images im skizzierten Sinne Rechnung tragen. Ein empirisches Indiz der Vorrangstellung ih- rer Image-Kommunikationen sind z.B. solche Inszenierungen, die die unter modernen Medienbedingungen zunehmend knapp werdende Ressource Aufmerksamkeit durch extreme Abweichungen vom Erwarteten und Üblichen attrahieren wollen. Die Wer- bungsproduzenten stellen diesbezüglich nämlich fest, daß die zur Steigerung von Auf-
• Visual aspects and physical location of the school are good – bad • Equipment of the school is modern – old • Study programme is difficult – easy • Innovation of the study programme is fast – slow • Range of extracurricular activities is large – poor • School climate is friendly – unfriendly • Children’s behaviour is appropriate – inappropriate • Success of graduates is high – low • Quality of the teaching staff is high – low • Management of the school is efficient – inefficient • Parental involvement is active – passive • Co-operation with the local community and employers is strong – weak • Partners’ relations and international relations are powerful – weak • Promotion of the school is well known – unknown To interpret and report the survey, creating a graphic presentation of the results of the questionnaire, in which each group of respondents is represented by its own line, is recommended. Results can be presented as a picture in which the average scores of each group of respondents are connected into one line. Each school image (view of a selected group of respondents) is represented by a vertical ‘line of means’ that summarises the average perception of the school. The result of each item depends not only on the means; it is necessary to ana- lyse the frequency of the respondent’s answers in each item of the partial scale. The frequency distribution is very important. ‘Because each image profile is a line of means, it does not reveal how variable the image is’ (Kotler, 2003, p. 567) Extreme values may mean that the image is highly specific or highly diffused.
the non-negativity of K, leads to processes related to backwards parabolic equations, which are known in to lack in general of a well-possednes theory. It is natural to expect an analogous property for their nonlocal counterpart. Backwards diffusion has been widely studied in previous literature. Its poten- tial applications on problems like the deblurring or sharpening of degraded imagery justify the efforts to overcome its inherent instability. The most widely used strategy in that direction, is to impose constraints at extrema that aim at enforcing a maximum–minimum principle. One example is the inverse diffusion filter of Osher and Rudin , which implements backward diffusion everywhere except at extrema, where the evolution is set to zero. Another example is the so-called forward-and-backward (FAB) diffusion of Gilboa et al. . It differs from the closely related Perona–Malik filter  by the fact that it uses negative diffusivities for a specific range of gradient magnitudes. However, at extrema where the gradient vanishes, it always avoids explosions by imposing forward diffusion. Another, less widely-used class of stabilisation attempts adds a fidelity term that prevents the back- ward evolution to move too far away from the original image  or from the average grey value of the desired range . In this case the range of the filtered image obviously depends on the weights of the fidelity and the backward diffusion term. Other examples of backwards processes are given in , where the authors show that the solution of a complex diffusion equa- tion combines properties of both forward and backward diffusion, and in  where the authors consider the flow of a nonconvex triple well potential. The main concern of this chapter will be the stabilisation of our nonlocal process without the non-negativity constraint.
at first glance, it can have a large impact on the simulation quality of the depth-of-field effect. To demonstrate this, we compare different forward op- erators in Figure 4.6. As one can see, the result of the forward operator of Aguet et al.  shown in Figure 4.6(c) is very close to simple 3-D convolu- tion (cf. Equation (4.10)) given in Figure 4.6(b). Differences are caused due to the discretisation required when embedding the pinhole image into the discrete 3-D volume according to Equation (4.9). Since both forward models perform convolutions with an unnormalised kernel h on strong depth changes, they produce bright overshoots followed by dark shadows. This behaviour is illustrated in Figure 4.5. In Figure 4.5(a) the slice taken out of the PSF consists of nearly a complete Gaussian caused by the blue surface and large part of a second Gaussian caused by the red surface. Hence, the integration weight of the slice exceeds 1 which leads to overshoots. In Figure 4.5(b) the slice only consists of approximately three-quarter of a Gaussian caused by the red surface. The rest is set to 0. Thus, the integration weight is lower than 1 which results in undershoots. Indeed, regarding Figure 4.6(c) and 4.6(b) respectively these local violations of the maximum-minimum principle on strong depth changes are violations of the model assumption and produce results that are not photorealistic. In contrast, comparing Figure 4.6(d) and (e) demonstrates that our normalised approach comes very close to the physically well-founded thin lens camera model, and allows to create realistic depth-of-field effects.
enhance the spatial resolution of the LR MS image at the expense of some loss in the spectral detail. Compared to the CS methods, MRA methods have better temporal coherence between the PAN and the enhanced MS image. MRA methods yield results with better spectral consistency at the expense of designing complex filters to mimic the reduction in the resolution by the sensors. Spatial details are injected into the MS data after a
However, the presence of social image concerns does not necessarily rely on the possibility to communicate. In fact, even without communication evidence for social image concerns has been found, such as the aversion of being perceived as egoistic or greedy (e.g. Ariely et al., 2009; Dana et al., 2006, 2007; G¨ uth et al., 1996; Koch and Normann, 2008; Tadelis, 2011). From a theoretical point of view, Tadelis (2011) proposes a model of “shame” inducing disutility of being perceived as a non-cooperator, in order to explain the effect he observes. 13 Besides the social disapproval of egoism, we are interested in the existence of another social norm which condemns promise breaking, thereby inducing additional social image concerns. Accordingly, we hypothesize that the effect of revelation on Roll rates is larger if subjects can communicate than without communication, indicating that the differential effect has to be due to an aversion to be regarded as a promise breaker. Hence, we compare the results of Com to the control treatment NoCom and state the following hypothesis. 14
Die Gesamtmittelwertgewichtungen zeigen keine bedeutenden Veränderungen der Imageprofile. Als beachtenswert kann jedoch erkannt werden, dass sich der durch- schnittliche Itemmittelwert des Imageprofil der Marke Samsung im Vergleich zur Ge- samtstichprobenerhebung in keiner Ausprägungsunterscheidung verändert. Hier können Rückschlüsse auf die Charakteristik von starken Marken als Interpretations- möglichkeit dieser Ergebnisse geschlossen werden. Starke Marken, zu denen auch die Marke Samsung eingeordnet werden kann, weisen sich aus schematheoretischer Per- spektive als ein sehr umfangreiches und tief strukturiertes Markenschema auf. Ihnen wird eine Selektion-und Filterfunktion zugesprochen, wodurch negative Informationen nur mit geringer Wahrscheinlichkeit wahrgenommen werden bzw. zur Akzeptanz füh- ren. Dadurch gelten starke Markenschemata als schwierig veränderbar, da sie sich durch ein stark verfestigtes Markenimage im Gedächtnis des Konsumenten darstellen. Es ist daher davon von auszugehen, dass die Probandengruppe mit negativer Einstel- lung zum Ambushing diese Art der Kommunikativen Maßnahme zwar nicht für guthei- ßen, jedoch die Marke Samsung auch nicht als Ambusher wahrnehmen, da die Probanden bei ihr auf ein stark verfestigtes Image zurückgreifen.
of registering all images at once. Furthermore, the distance mea- sure is capable of comparing structural information across all im- ages, based on the chosen features. The alignment based on these features is based on a geometric idea. The intensity changes of an image is qualiﬁed as edges or jumps from a dark to a lighter region or vice versa. The gradients describe the direction of the jump as well as its height. Since the height of the jump is dif- ferent in diﬀerent modalities, a normalization of the height is considered. After normalization, the idea is to align the gradient directions. Since the same or at least equal tissue is shown on the given sections, the edges should be in corresponding spots. Im- ages are not ﬁxed and the alignment of each image is dependent of all features of all other images. For further technical details on two-image approaches and other groupwise approaches, ref- erence is made to, for example, refs. [7, 9].
The software is designed primarily for utilization in On-Orbit Servicing tasks, where- for example- a servicer spacecraft approaches an uncooperative client spacecraft, which can not aid in the process in any way as it is assumed to be completely passive. Image processing is used for navigating to the client spacecraft. In this specific scenario, it is vital to obtain accurate distance and bearing information until, in the last few meters, all six degrees of freedom are needed to be known. The smaller the distance between the spacecraft, the more accurate pose estimates are required.
But such a limited number of points still can be used for a coarse image registration. The concept of a two-stage registration is fairly spread in optical image registration. At a reduced resolution some good points for a coarse registration are extracted and e.g. the coefficients of an affine transformation are calculated. Then, a fine matching takes place. Such a registration strategy, i.e. a two-stage registration, can be highly recommended for matching of SAR data. First of all, this technique is successfully used in optical data registration procedures. Secondly, it has the great advantage for SAR images that in very coarse resolution the speckle almost cancels out. And thirdly, for SAR data does a large amount of fine registration techniques from InSAR exist, which necessarily need good coarse registration.
Cooperation among interacting partners is essential for economic success in many situ- ations, as joint value creation often exceeds individual achievements. These situations become challenging as soon as cooperation cannot be contractually enforced, but relies on mutual trust by the interacting partners. Among a large literature focusing on how to improve cooperation, various experimental studies show that communication can be an effective tool to enhance it (see, e.g. Bochet and Putterman, 2009; Cooper et al., 1992; Ellingsen and Johannesson, 2004). While several articles analyze whether cheap talk can be effective and how this depends on the communication protocol and the game structure (see for instance Blume and Ortmann, 2007; Camera et. al., 2011; Ellingsen and ¨ Ostling, 2010; Kriss et al., 2011; Mohlin and Johanneson, 2008), we contribute to the literature focusing on why individuals stick to a commitment, given that rationality predicts a deviating behavior. In particular, we analyze whether and to what extent social image concerns motivate people to stick to a given promise. More precisely, as breaking a promise is deemed negative in society, avoiding the image of being a promise breaker might induce individuals to keep their word. Consequently, we study whether an individual is more likely to act in line with a given promise if its violation is more obvious to its receiver.