Janis Fehr

3D rotation invariant local binary patterns

We present a novel method for the fast computation of rotation invariant ¿local binary patterns¿ (LBP) on 3D volume data. Unlike a previous publication on 3D LBP, this new approach is not limited to ¿uniform patterns¿, providing a real 3D extension of the standard and rotation invariant LBP. We evaluate our methods in the context of 3D texture analysis of biological data.

Computing range flow from multi-modal Kinect data

In this paper, we present a framework for range flow estimation from Microsofts multi-modal imaging device Kinect. We address all essential stages of the flow computation process, starting from the calibration of the Kinect, over the alignment of the range and color channels, to the introduction of a novel multi-modal range flow algorithm which is robust against typical (technology dependent) range estimation artifacts.

Voxel-wise gray scale invariants for simultaneous segmentation and classification

3D volumetric microscopical techniques (e.g. confocal laser scanning microscopy) have become a standard tool in biomedical applications to record three-dimensional objects with highly anisotropic morphology. To analyze these data in high-throughput experiments, reliable, easy to use and generally applicable pattern recognition tools are required. The major problem of nearly all existing applications is their high specialization to exact one problem, and the their time-consuming adaption to new problems, that has to be done by pattern recognition experts. We therefore search for a tool that can be adapted to new problems just by an interactive training process. Our main idea is therefore to combine object segmentation and recognition into one step by computing voxel-wise gray scale invariants (using nonlinear kernel functions and Haar-integration) on the volumetric multi-channel data set and classify each voxel using support vector machines. After the selection of an appropriate set of nonlinear kernel functions (which allows to integrate previous knowledge, but still needs some expertise), this approach allows a biologist to adapt the recognition system for his problem just by interactively selecting several voxels as training points for each class of objects. Based on these points the classification result is computed and the biologist may refine it by selecting additional training points until the result meets his needs. In this paper we present the theoretical background and a fast approximative algorithm using FFTs for computing Haar-integrals over the very rich class of nonlinear 3-point-kernel functions. The approximation still fulfils the invariance conditions. The experimental application for the recognition of different cell cores of the chorioallantoic membrane is presented in the accompanying paper [1] and in the technical report [2]

Pericytes in the mature chorioallantoic membrane capillary plexus contain desmin and α-smooth muscle actin: relevance for non-sprouting angiogenesis

The chicken chorioallantoic membrane (CAM) is a frequently used tissue for studying vascular growth and remodeling, notably non-sprouting angiogenesis by formation of transluminal pillars. Vascular pericytes have received increasing attention in the field of angiogenesis research and appear important for pillar growth. Our earlier observation that desmin (DES), but not α-smooth muscle actin (αSMA) was expressed in pericytes of the mature CAM capillary plexus after E12 was confirmed by others. However, in different species or vascular beds, either marker or both have been used to identify pericytes, raising the questions if (1) expression of these cytoskeletal proteins really was mutually exclusive; or (2) different types of pericytes existed in the same vascular bed. Using triple labeling with fluorochrome-conjugated markers Sambucus nigra agglutinin, DES or αSMA, and DNA-specific YoPro-1, we report here for the first time a delicate filamentous, circumferentially oriented αSMA pattern in periendothelial cells of the mature CAM capillary plexus, quite different from the coarser, axially oriented DES pattern. A new method for automatic classification of DNA-staining pattern was applied to compare nuclei of DES- or αSMA-positive cells. It predicted colocalisation of both proteins in most capillary pericytes, which was confirmed by double immunostaining for DES and αSMA. We conclude that (1) in contrast to published work, DES and αSMA are not mutually exclusive in most pericytes; (2) different types of pericytes may co-exist in the same vascular bed; (3) on average, one pericyte is associated with two transluminal pillars; (4) a novel imaging modality may be useful for cell identification in angiogenesis research and elsewhere.

A Bag of Features Approach for 3D Shape Retrieval

In this paper, we present an adaptation the Bag of Features (BoF) concept to 3D shape retrieval problems. The BoF approach has recently become one of the most popular methods in 2D image retrieval. We extent this approach from 2D images to 3D shapes. Following the BoF outline, we address the necessary modifications for the 3D extension and present novel solutions for the parameterization of 3D patches, a 3D rotation invariant similarity measure for these patches and a method for the codebook generation. We experimentally evaluate the performance of our methods on the Princeton Shape Benchmark .

Self-learning segmentation and classification of cell-nuclei in 3D volumetric data using voxel-wise gray scale invariants

We introduce and discuss a new method for segmentation and classification of cells from 3D tissue probes. The anisotropic 3D volumetric data of fluorescent marked cell nuclei is recorded by a confocal laser scanning microscope (LSM). Voxel-wise gray scale features (see accompaning paper [1][2]) ), invariant towards 3D rotation of its neighborhood, are extracted from the original data by integrating over the 3D rotation group with non-linear kernels. In an interactive process, support-vector machine models are trained for each cell type using user relevance feedback. With this reference database at hand, segmentation and classification can be achieved in one step, simply by classifying each voxel and performing a connected component labelling, automatically without further human interaction. This general approach easily allows adoption of other cell types or tissue structures just by adding new training samples and re-training the model. Experiments with datasets from chicken chorioallantoic membrane show encouraging results.

Local Rotation Invariant Patch Descriptors for 3D Vector Fields

In this paper, we present two novel methods for the fast computation of local rotation invariant patch descriptors for 3D vectorial data. Patch based algorithms have recently become very popular approach for a wide range of 2D computer vision problems. Our local rotation invariant patch descriptors allow an extension of these methods to 3D vector fields. Our approaches are based on a harmonic representation for local spherical 3D vector field patches, which enables us to derive fast algorithms for the computation of rotation invariant power spectrum and bispectrum feature descriptors of such patches.

Fast Support Vector Machine Classification using linear SVMs

We propose a classification method based on a decision tree whose nodes consist of linear support vector machines (SVMs). Each node defines a decision hyperplane that classifies part of the feature space. For large classification problems (with many support vectors (SVs)) it has the advantage that the classification time does not depend on the number of SVs. Here, the classification of a new sample can be calculated by the dot product with the orthogonal vector of each hyperplane. The number of nodes in the tree has shown to be much smaller than the number of SVs in a non-linear SVM, thus, a significant speedup in classification time can be achieved. For non-linear separable problems, the trivial solution (zero vector) of a linear SVM is analyzed and a new formulation of the optimization problem is given to avoid it

Fast Support Vector Machine Classification of Very Large Datasets

In many classification applications, Support Vector Machines (SVMs) have proven to be highly performing and easy to handle classifiers with very good generalization abilities. However, one drawback of the SVM is its rather high classification complexity which scales linearly with the number of Support Vectors (SVs). This is due to the fact that for the classification of one sample, the kernel function has to be evaluated for all SVs. To speed up classification, different approaches have been published, most which of try to reduce the number of SVs. In our work, which is especially suitable for very large datasets, we follow a different approach: as we showed in (Zapien et al. 2006), it is effectively possible to approximate large SVM problems by decomposing the original problem into linear subproblems, where each subproblem can be evaluated in Ω(1). This approach is especially successful, when the assumption holds that a large classification problem can be split into mainly easy and only a few hard subproblems. On standard benchmark datasets, this approach achieved great speedups while suffering only sightly in terms of classification accuracy and generalization ability. In this contribution, we extend the methods introduced in (Zapien et al. 2006) using not only linear, but also non-linear subproblems for the decomposition of the original problem which further increases the classification performance with only a little loss in terms of speed. An implementation of our method is available in (Ronneberger and et al.) Due to page limitations, we had to move some of theoretic details (e.g. proofs) and extensive experimental results to a technical report (Zapien et al. 2007).

Fast Generalized Belief Propagation for MAP Estimation on 2D and 3D Grid-Like Markov Random Fields

In this paper, we present two novel speed-up techniques for deterministic inference on Markov random fields (MRF) via generalized belief propagation (GBP). Both methods require the MRF to have a grid-like graph structure, as it is generally encountered in 2D and 3D image processing applications, e.g. in image filtering, restoration or segmentation. First, we propose a caching method that significantly reduces the number of multiplications during GBP inference. And second, we introduce a speed-up for computing the MAP estimate of GBP cluster messages by presorting its factors and limiting the number of possible combinations. Experimental results suggest that the first technique improves the GBP complexity by roughly factor 10, whereas the acceleration for the second technique is linear in the number of possible labels. Both techniques can be used simultaneously.