Journal articles

1. Bio-optical Modelling Combined with Remote Sensing to Assess Water Quality
   Authors: Ammenberg-Philipson, P.; Flink, P.; Pierson, D. (1); Lindell, T.; Strömbeck, N. (1)
   (1) Dept. of Evolutionary Biology, Limnology, UU
   Journal: International Journal of Remote Sensing, 23(8):1621-1638, 2002
   Abstract: A simple bio-optical model, with parameter values derived from measurements of the inherent optical properties (IOPs) and optically active substances that are known to influence the IOPs, has been developed. A large historical dataset of measurements of the concentration of chlorophyll a and phaeophytine a (Chl), suspended particulate inorganic material (SPIM) and the absorption coefficient of Coloured Dissolved Organic Matter (CDOM), spanning more than 25 years, has been used together with the model to develop algorithms for the retrieval of these water quality parameters, for a site in Lake Malaren, Sweden. The model takes as input the optically active substances and outputs a reflectance spectrum just above the water surface. From the modelled reflectance, algorithms were derived for Chl, SPIM and CDOM absorption at 420 nm. The algorithms were applied to atmospherically corrected remote sensing data, which were collected by the Compact Airborne Spectrographic Imager, CASI. The radiative transfer code 6S was used for the atmospheric correction of the data. Distribution maps for the three retrieved parameters were constructed and Chl and SPIM were validated by continuous field measurements of fluorescence and beam attenuation. The continuous data were calibrated with water analysis results from nine water samples. The time lag between the image acquisition and the ground data measurements was never more than 3 hours. Even though the model parameter values were collected at different times from that of the CASI over-flight, and from a larger geographic region of Lake Malaren than that used for the CASI measurements, the independently developed algorithms predicted the concentrations of the optically active substances within a reasonable level of accuracy, allowing spatial variations in the substances to be predicted.

2. Slice Based Digital Volume Assembly of a Small Paper Sample
   Authors: Aronsson, M.; Henningsson, O. (1); Sävborg, Ö. (1)
   (1) StoraEnso Research, Falun
   Journal: Nordic Pulp and Paper Research Journal, 17(1), 2002
   Abstract: Digital volume images can be created by assembling a stack of 2D images. By using a microtome for slicing, a Scanning Electron Microscope for imaging and digital analysis tools, we were able to create a small digital volume from a paper sample of Duplex-board. Imaging the surface of the embedding rather than the cut-off slices, was crucial in minimizing geometrical distortions. The use of reference threads ensured a good registration with small errors and reasonable effort. For visualization purposes, we used a surface renderer based on the Marching Cube algorithm. Although the reconstruction process is time consuming, it is a viable methods for creating volume images of paper with micrometer resolution.

3. A Segmentation Technique to Determine Fat Content in NMR Images of Beef Meat
   Authors: Ballerini, L.; Högberg, A. (1); Borgefors. G.; Bylund, A.-C. (1); Lindgård, A. (1); Lundström, K. (1); Rakotonirainy, O. (2); Soussi, B. (2)
   (1) Dept. of Food Science, SLU, Uppsala;
   (2)  Wallenberg Laboratory, Sahlgrenska University Hospital, Göteborg
   Journal: IEEE Transactions on Nuclear Science, 49(1):195-199, 2002
   Abstract: The world of meat faces a permanent need for new methods of meat quality evaluation. Recent advances in the area of computer and video processing have created new ways to monitor quality in the food industry. In this paper we describe an image processing technique to determine fat content in beef meat. To achieve this, NMR (Nuclear Magnetic Resonance) images of beef meat have been used. The inherent advantages of NMR images are many. Chief among these are unprecedented contrasts between the various structures present in meat, such as muscle, fat, and connective tissue. Moreover, the three-dimensional nature of the NMR method allows the analysis of isolated cross-sectional slices of the meat and the measure of volumetric content of fat, and it is not limited to measurements of the superficially visible fat. We propose a segmentation algorithm for the detection of fat and a filtering technique to remove intensity inhomogeneities in NMR images, caused by non-uniformities of magnetic field during acquisition. Measurements have been successfully correlated with chemical analysis and digital photography. We also propose a method to quantify the distribution of fat. Our results show that NMR technique is a promising non-invasive method to determine fat content in meat.

4. Individual Tree-based Species Classification in High Spatial Resolution Aerial Images of Forests using Fuzzy Sets
   Author: Brandtberg, T.
   Journal: Fuzzy Sets and Systems, 132(3):371-387, 2002
   Abstract: This paper presents an application of fuzzy set theory for classification of individual tree crowns into species groups, in high spatial resolution colour infrared aerial photographs. In this type of digital image, the trees are visible as individual objects. The number of individuals to classify might be very large in the acquired set of photographs, but the applied grade of membership (GoM) model, which this paper focuses on, is suitable for dealing with large datasets. The extent of each tree crown in the image is defined using a previously published procedure. Based on colour information (hue), an optimal fuzzy thresholding technique divides the tree crown universal set into a dominant set and its minor complement. Nine different features of each image object are estimated, and transformed using principal component analysis (PCA). The first three or four PCs are subsequently used in the GoM model. Furthermore, the concept of fuzzy relation is applied to one of the descriptors: to predict a centroid of the star-shaped pattern of Norway spruce. The GoM model needs initial membership values, which are estimated using an unsupervised fuzzy clustering approach of small subareas (branches in the tree crowns) and their corresponding digital numbers in each colour band (RGB-images). The complete classification system comprises three independent components: decisions on coniferous/deciduous, Scots pine/Norway spruce, and Birch/Aspen. The accuracies (ground patches excluded), using the supervised GoM model with crossvalidation, are 87%, 76%, and 79%, respectively. The accuracy for the compounded system is 67%.

5. Using Weighted Fixed Neural Networks for Unsupervised Fuzzy Clustering
   Author: Hamid Muhammed, H.
   Journal: International Journal of Neural Systems, 12(6):425-434, 2002
   Abstract: A novel algorithm for unsupervised fuzzy clustering is introduced. The algorithm uses a so-called Weighted Fixed Neural Networks (WFNN) to store important and useful information about the topological relations in a given data set. The algorithm produces a weighted connected net, of weighted nodes connected by weighted edges, which reflects and preserves the topology of the input data set. The weights of the nodes and the edges in the resulting net are proportional to the local densities of data samples in input space. The connectedness of the net can be changed, and the higher the connectedness of the net is chosen, the fuzzier the system becomes. The new algorithm is computationally efficient when compared to other existing methods for clustering multi-dimensional data, such as colour images.

6. Tripeptide Interference with Human Immunodeficiency Virus Type 1 Morphogenesis
   Authors: Höglund, S. (1); Su, J. (2); Sundin Reneby, S. (1); Végvári, Á. (1); Hjertén, S. (1); Sintorn, I.-M.; Foster, H. (1); Wu, Y.-P. (2); Nyström, I.; Vahlne, A. (2)
   (1) HIV structure group, Dept. of Biochemistry, UU
   (2) Div. of Clinical Virology, The Karolinska Institut, Huddinge University Hospital, Stockholm
   Journal: Antimicrobial Agents and Chemotherapy, 46(11):3597-3605, 2002
   Abstract: Capsid assembly during virus replication is a potential target for antiviral therapy. The Gag polyprotein is the main structural component of retroviral particles, and in human immunodeficiency virus type 1 (HIV-1), it contains the sequences for the matrix, capsid, nucleocapsid, and several small polypeptides. Here, we report that at a concentration of 100 M, 7 of 83 tripeptide amides from the carboxyl-terminal sequence of the HIV-1 capsid protein p24 suppressed HIV-1 replication (>80%). The three most potent tripeptides, glycyl-prolyl-glycine-amide (GPG-NH2), alanyl-leucyl-glycine-amide (ALG-NH2), and arginyl-glutaminyl-glycine-amide (RQG-NH2), were found to interact with p24. With electron microscopy, disarranged core structures of HIV-1 progeny were extensively observed when the cells were treated with GPG-NH2 and ALG-NH2. Furthermore, nodular structures of approximately the same size as the broad end of HIV-1 conical capsids were observed at the plasma membranes of treated cells only, possibly indicating an arrest of the budding process. Corresponding tripeptides with nonamidated carboxyl termini were not biologically active and did not interact with p24.

7. A New Shape Descriptor for Surfaces in 3D Images
   Authors: Sanniti di Baja, G. (1); Svensson, S..
   (1) Istituto di Cibernetica, National Research Council of Italy (CNR), Arco Felice (Napoli), Italy
   Journal: Pattern Recognition Letters, 23(6):703-711, 2002
   Abstract: We introduce a linear shape descriptor for (open) surfaces in 3D images. To extract the shape descriptor, the border of the surface is first identified. Then, the distance transform of the surface is computed, where each voxel in the surface is labelled with the minimum distance to its closest border voxel. On the distance transform, the centres of the maximal geodesic discs (CMGDs) are detected. These voxels are suitably linked to each other by growing paths in the direction of the steepest gradient, to finally obtain the linear shape descriptor of the surface. The shape descriptor can be extracted from any open surface-like object, i.e., an object with thickness at most two-voxel.

8. Digital Distance Transforms in 3D Images Using Information from Neighbourhoods up to 5×5×5
   Authors: Svensson, S.; Borgefors, G.
   Journal: Computer Vision and Image Understanding, 88(1):24-53, 2002
   Abstract: 3D distance image, or a distance transform, is an image where each feature voxel is labeled with the distance to its closest nonfeature voxel. Distance transforms are useful for many binary (shape) image analysis tasks. The distance transform can be computed by propagating local distance information between neighboring voxels. In a weighted distance transform, the local distances are optimized to make the distance transform more stable under rotation. We present results from optimization for 3D images when using from one to six local distances, all in the 5×5×5 neighbourhood of a voxel.

9. Distance Transforms in 3D using Four Different Weights
   Authors: Svensson, S.; Borgefors, G.
   Journal: Pattern Recognition Letters, 23(12):1407-1418, 2002
   Abstract: Digital distance transformations provide helpful tools for representation and description of object shape in digital images. The resulting distance transforms should be stable under trans ation and rotation. To this end, the Euclidean distance is approximated. We present results for distance transforms for 3D images, where the four weights, or local distances, are used, the three weights from the 3×3×3 neighbourhood together with one weight from the outer part of the 5×5×5 neighbourhood.

10. Curve Skeletonization of Surface-Like Objects in 3D Images Guided by Voxel Classification
    Authors: Svensson, S.; Nyström, I.; Sanniti di Baja, G. (1)
    (1) Istituto di Cibernetica, National Research Council of Italy (CNR), Arco Felice (Napoli), Italy
    Journal: Pattern Recognition Letters, 23(12):1419-1426, 2002
    Abstract: Skeletonization is a way to reduce dimensionality of digital objects. Here, we present an algorithm that computes the curve skeleton of a surface-like object in a 3D image, i.e., an object that in one of the three dimensions is at most two-voxel thick. A surface-like object consists of surfaces and curves crossing each other. Its curve skeleton is a 1D set centred within the surface-like object and with preserved topological properties. It can be useful to achieve a qualitative shape representation of the object with reduced dimensionality. The basic idea behind our algorithm is to detect the curves and the junctions between different surfaces and prevent their removal as they retain the most significant shape representation.

11. Using Distance Transforms to Decompose 3D Discrete Objects
    Authors: Svensson, S.; Sanniti di Baja, G. (1)
    (1) Istituto di Cibernetica, National Research Council of Italy (CNR), Arco Felice (Napoli), Italy
    Journal: Image and Vision Computing, 20(8):529-540, 2002
    Abstract: Object decomposition into simpler parts greatly diminishes the complexity of a recognition task. In this paper, we present a method to decompose a 3D discrete object into nearly convex or elongated parts. Object decomposition is guided by the distance transform (DT). Significant voxels in DT are identified and grouped into seeds. These are used to originate the parts of the object by applying the reverse and the constrained distance transformations. Criteria for merging less significant parts and obtaining a perceptually meaningful decomposition are also given. This approach is likely to be of interest in future applications due to the increasing number and the decreasing cost of devices for volume image acquisition.

12. Segmentation with Gray-Scale Connectedness can Separate Arteries and Veins in MRA
    Authors: Tizon, X.; Smedby, Ö. (1)
    (1) Dept. of Medicine and Care, Linköping University Hospital
    Journal: Journal of Magnetic Resonance Imaging, 15(4):438-445, 2002
    Abstract:
    Purpose: To describe and present some preliminary results for a novel algorithm for segmentation with gray-scale connectedness as a means to separate arteries and veins in magnetic resonance angiography (MRA).
    Materials and Methods: The proposed algorithm, SeparaSeed, uses the gray-scale degree of connectedness as a tool to find the zone surrounding each vessel, in order to split the original volume into its different vessel components. In contrast to traditional segmentation methods, no gray-scale information is lost in the process. The segmentation is performed in one step, resulting in a partition of the initial volume into a chosen number of regions of interest (ROIs). Finally, visualization is achieved by projecting the 3D vessel trees to 2D using the common maximum intensity projection (MIP). The algorithm was tested in two MRA data sets of the vessels of the pelvis acquired after injection of an intravascular contrast agent and in one data set of the vessels of the neck with gadolinium.
    Results: In all data sets, a large proportion of the venous signal was removed while preserving that of the arteries, thus improving visualization of the relevant vessels.
    Conclusion: Separation of arteries and veins is feasible with the proposed algorithm with a moderate amount of interaction.