Gustaf Kylberg

Evaluation of noise robustness for local binary pattern descriptors in texture classification

Local binary pattern (LBP) operators have become commonly used texture descriptors in recent years. Several new LBP-based descriptors have been proposed, of which some aim at improving robustness to noise. To do this, the thresholding and encoding schemes used in the descriptors are modified. In this article, the robustness to noise for the eight following LBP-based descriptors are evaluated; improved LBP, median binary patterns (MBP), local ternary patterns (LTP), improved LTP (ILTP), local quinary patterns, robust LBP, and fuzzy LBP (FLBP). To put their performance into perspective they are compared to three well-known reference descriptors; the classic LBP, Gabor filter banks (GF), and standard descriptors derived from gray-level co-occurrence matrices. In addition, a roughly five times faster implementation of the FLBP descriptor is presented, and a new descriptor which we call shift LBP is introduced as an even faster approximation to the FLBP. The texture descriptors are compared and evaluated on six texture datasets; Brodatz, KTH-TIPS2b, Kylberg, Mondial Marmi, UIUC, and a Virus texture dataset. After optimizing all parameters for each dataset the descriptors are evaluated under increasing levels of additive Gaussian white noise. The discriminating power of the texture descriptors is assessed using tenfolded cross-validation of a nearest neighbor classifier. The results show that several of the descriptors perform well at low levels of noise while they all suffer, to different degrees, from higher levels of introduced noise. In our tests, ILTP and FLBP show an overall good performance on several datasets. The GF are often very noise robust compared to the LBP-family under moderate to high levels of noise but not necessarily the best descriptor under low levels of added noise. In our tests, MBP is neither a good texture descriptor nor stable to noise.

Virus texture analysis using local binary patterns and radial density profiles

We investigate the discriminant power of two local and two global texture measures on virus images. The viruses are imaged using negative stain transmission electron microscopy. Local binary patterns and a multi scale extension are compared to radial density profiles in the spatial domain and in the Fourier domain. To assess the discriminant potential of the texture measures a Random Forest classifier is used. Our analysis shows that the multi scale extension performs better than the standard local binary patterns and that radial density profiles in comparison is a rather poor virus texture discriminating measure. Furthermore, we show that the multi scale extension and the profiles in Fourier domain are both good texture measures and that they complement each other well, that is, they seem to detect different texture properties. Combining the two, hence, improves the discrimination between virus textures.

Segmentation of virus particle candidates in transmission electron microscopy images

In this paper, we present an automatic segmentation method that detects virus particles of various shapes in transmission electron microscopy images. The method is based on a statistical analysis of local neighbourhoods of all the pixels in the image followed by an object width discrimination and finally, for elongated objects, a border refinement step. It requires only one input parameter, the approximate width of the virus particles searched for. The proposed method is evaluated on a large number of viruses. It successfully segments viruses regardless of shape, from polyhedral to highly pleomorphic.

Local intensity and PCA based detection of virus particle candidates in transmission electron microscopy images

We present a general method using local intensity information and PCA to detect objects characterized only by that they differ from their surroundings. We apply our method to the problem of automatically detecting virus particle candidates in transmission electron microscopy images. Viruses have very different shapes and sizes, many species are spherical whereas others are highly pleomorphic. To detect any kind of virus particles in electron microscopy images it is therefore necessary to use a method not restricted to detection of a specific shape. The method proposed here uses only one input parameter, the approximate virus thickness, which is a conserved feature within a virus species. It is capable to detect virus particles of very varying shapes. Results on images with highly textured background of several different virus species are presented.

Towards Automated TEM for Virus Diagnostics: Segmentation of Grid Squares and Detection of Regions of Interest

When searching for viruses in an electron microscope the sample grid constitutes an enormous search area. Here, we present methods for automating the image acquisition process for an automatic virus diagnostic application. The methods constitute a multi resolution approach where we first identify the grid squares and rate individual grid squares based on content in a grid overview image and then detect regions of interest in higher resolution images of good grid squares. Our methods are designed to mimic the actions of a virus TEM expert manually navigating the microscope and they are also compared to the expert’s performance. Integrating the proposed methods with the microscope would reduce the search area by more than 99.99 % and it would also remove the need for an expert to perform the virus search by the microscope.

Virus Recognition Based on Local Texture

To detect and identify viruses in electron microscopy images is crucial in certain clinical emergency situations. It is currently a highly manual task, requiring an expert sitting at the microscope to perform the analysis visually. Here we focus on and investigate one aspect towards automating the virus diagnostic task, namely recognizing the virus type based on their texture once possible virus objects have been segmented. We show that by using only local texture descriptors we achieve a classification rate of almost 89% on texture patches from 15 different virus types and a debris (false object) class. We compare and combine 5 different types of local texture descriptors and show that by combining the different types a lower classification error is achieved. We use a Random Forest Classifier and compare two approaches for feature selection.

A Fast Fourier based Feature Descriptor and a Cascade Nearest Neighbour Search with an Efficient Matching Pipeline for Mosaicing of Microscopy Images

Automatic mosaicing is an important image processing application and we propose several improvements and simplifications to the image registration pipeline used in microscopy to automatically construct large images of whole specimen samples from a series of images. First of all we propose a feature descriptor based on the amplitude of a few elements of the Fourier transform, which makes it fast to compute and that can be used for any image matching and registration applications where scale and rotation invariance is not needed. Secondly, we propose a cascade matching approach that will reduce the time for the nearest neighbour search considerably, making it almost independent on feature vector length. Moreover, several improvements are proposed that will speed up the whole matching process. These are: faster interest point detection, a regular sampling strategy and a deterministic false positive removal procedure that finds the transformation. All steps of the improved pipeline are explained and the results comparative experiments are presented.

An Efficient Descriptor Based on Radial Line Integration for Fast Non-invariant Matching and Registration of Microscopy Images.

Descriptors such as SURF and SIFT contain a framework for handling rotation and scale invariance, which generally is not needed when registration and stitching of images in microscopy is the focus. Instead speed and efficiency are more important factors. We propose a descriptor that performs very well for these criteria, which is based on the idea of radial line integration. The result is a descriptor that outperforms both SURF and SIFT when it comes to speed and the number of inliers, even for rather short descriptors.

Automated Image Analysis Using a Tabletop Low Voltage TEM

The development of technologies using nanoparticles has been significantly increasing over the past decade, with applications in a broad scope of fields, as well in material sciences as in life sciences, and the demand on particle characterization has been growing accordingly. Amongst the methods available for particle characterization, Electron Microscopic technologies in contrast to indirect methods like Dynamic Light Scattering and Nano Tracking Analysis result in high resolution images that provide morphological information and particle identification. EM however requires an experienced operator and can be time and resource consuming. A method allowing to simplify the analysis process with an automated image acquisition of samples as well as an automatic image analysis on a low-voltage electron microscope is proposed herein.