Mariusz Paradowski

Local Keypoints and Global Affine Geometry: Triangles and Ellipses for Image Fragment Matching

Image matching and retrieval is one of the most important areas of computer vision. The key objective of image matching is detection of near-duplicate images. This chapter discusses an extension of this concept, namely, the retrieval of near-duplicate image fragments. We assume no a’priori information about visual contents of those fragments. The number of such fragments in an image is also unknown. Therefore, we address the problem and propose the solution based purely on visual characteristics of image fragments The method combines two techniques: a local image analysis and a global geometry synthesis. In the former stage, we analyze low-level image characteristics, such as local intensity gradients or local shape approximations. In the latter stage, we formulate global geometrical hypotheses about the image contents and verify them using a probabilistic framework.

Capillary Abnormalities Detection Using Vessel Thickness and Curvature Analysis

The growing importance of nail-fold capillaroscopy imaging as a diagnostic tool in medicine increases the need to automate this process. One of the most important markers in capillaroscopy is capillary thickness. On this basis capillaries may be divided into three separate categories: healthy , capillaries with increased loops and megacapillaries . In the paper we describe the problem of capillary thickness analysis automation. First, data is extracted from a segmented capillary image. Then feature vectors are constructed. They are given as an input for capillary classification method. We applied different classifiers in the experiments. The best achieved accuracy reaches 97%, which can be considered as very high and satisfying.

Resulted word counts optimization-A new approach for better automatic image annotation

One of major problems in image auto-annotation is the difference between the expected word counts vector and the resulted word counts vector. This paper presents a new approach to automatic image annotation-an algorithm called resulted word counts optimizer which is an extension to existing methods. An ideal annotator is defined in terms of recall quality measure. On the basis of the ideal annotator an optimization criterion is defined. It allows to reduce the difference between resulted and expected word counts vectors. The proposed algorithm can be used with various image auto-annotation algorithms because its generic nature. Additionally, it does not increase the computational complexity of the original annotation method processing phase. It changes output word probabilities according to a pre-calculated vector of correction coefficients.

Multiple Class Machine Learning Approach for an Image Auto-Annotation Problem

Image auto-annotation problem becomes more and more popular research topic. Possible applications of auto-annotation methods range from Internet search engines to medical analysis software. The important aspect is that efficient image auto-annotation systems can eliminate the need of annotating huge image collections manually, which is the only solution today. Most of methods available in the literature do not use supervised machine learning as the key component. Recent researches show that supervised machine learning can successfully compete with existing approaches. This paper presents a novel image auto-annotation algorithm based of supervised machine learning with the use of C4.5 classifiers

Capillary Blood Vessel Tortuosity Measurement Using Graph Analysis

Capillaroscopy is a branch of medicine which allows to diagnose various kinds of rheumatic diseases on the basis of observation of visual properties of nail-fold capillaries. Capillaries are tiny blood vessels of various shapes and sizes. Blood vessel tortuosity is one of medical signs. The paper presents a novel blood vessel tortuosity measure designed for capillary analysis. It represents the vessel as a graph and utilizes non-directional and directional traversal algorithms.

Capillaroscopy Image Analysis as an Automatic Image Annotation Problem

Image auto-annotation methods perform the description previously unseen images with a set of words, based on previously shown examples. This is a relatively new paradigm in image processing research domain. Capillaroscopy is a branch of medicine focusing on analysis of vascular changes. Among others, it is a very helpful tool in diagnosis of many rheumatic diseases. The paper presents a preliminary research on usage of image auto-annotation methods for the computer-aided diagnosis from capillaroscopy images.

Intelligent Techniques in Personalization of Learning in e-Learning Systems

This chapter contains an overview of intelligent techniques that can be applied in different stages of e-learning systems to achieve personalization. It describes examples of their application to various e-learning platforms to create profiles of learners and to define learning path. The typical approach to obtain learner’s profile is the usage one of the clustering methods, such as: the simple k-means, Self Organizing Map, hierarchical clustering or fuzzy clustering. Classification methods like: C4.5 or C.5, k-Nearest Neighbor and Naive Bayes are also useful, but they need to define classes and training patterns by an expert. In contrary, clustering is unsupervised learning method and the categories are discovered by the method itself. The recommending system is responsible for proposing individual learning path for each learner. The most popular approach is an application of the Aprori method which searches for association rules. However, it seems that it is rather inefficient method when the number of data to process is huge. Other methods and models that can be useful for knowledge representation are also discussed. Recommending systems are mainly built as a knowledge based. Most of them are implemented as rule based systems. An interesting approach implementing cased based reasoning paradigm to recommend learning path is described as well. The end of the chapter contains a critical discussion of existing solutions and suggests possible research in this field.

Avascular area detection in nailfold capillary images

Automation of nailfold capillary image analysis is a new research idea. There are many important aspects of the capillary image analysis: capillary thickness, shape, distribution and density. We focus on two last aspects—distribution and density of capillaries. The paper presents an approach to avascular areas detection. An avascular area on the image is an area with capillary loss. It is an important medical sign. Capillary loss is one of the most characteristic features of systemic sclerosis. Vascular abnormalities observed in nailfold capillaries appear earlier in the course of the disease than at other sites. The proposed approach uses a variety of pattern recognition techniques, including: histogram analysis and classification.

Visual similarity issues in face recognition

The paper discusses several issues of visual similarity in face detection and recognition. Using a straightforward concept of keypoint correspondences, a method is proposed to formalise the subjective impressions of |similar faces|, |similar eyes|, |similar chins|, etc. The method exploits the mechanism of affine near-duplicate fragment detection originally proposed for visual information retrieval. It is shown that using such a method, a simple and relatively reliable face detection/identification systems can be build without any model (or training) of human faces, which can work with images containing multiple faces shown on random backgrounds. Additionally, it is proposed how the same approach can be used to optimise databases of face images and to identify individuals who are at higher risks of mistaken face identification.

Detection of Image Fragments Related by Affine Transforms: Matching Triangles and Ellipses

Visual information retrieval systems are often constructed upon the notion of image similarity. The concept of image similarity may be defined in many ways: from a pure visual level, where we seek identical images, to a semantic level related with human perception the image. In our research we address the first approach, we explore topics of image matching (image alignment), however in terms of image fragments. The goal of image fragment matching is to find similar parts of two images, without a given model of particular objects present on images. It is also assumed that the number of similar objects (image fragments) is not known. In this paper we present a novel method for image fragment matching. It uses two ellipse pairs as an elementary object for image geometry reconstruction. The method is an extension of the previously proposed approach based on triangles. We have decided to replace triangles with a different geometrical structure to reduce computational complexity from O(n^3) to O(n^2), where n is the number of coherent key regions. We discuss and compare both matching methods both in terms of quality and processing efficiency.