Dariusz Frejlichowski

An experimental comparison of seven shape descriptors in the general shape analysis problem

The general shape analysis is a problem similar to the recognition or retrieval of shapes. The most important difference is that the processed object does not have to belong to a base class, but usually is only similar to the template representing the class. The most general information about a shape is here concluded, i.e. how round, elliptical, triangular, etc. it is. Such a problem can occur in applications with few general base classes, e.g. in pre-classification or the assignment of stamps extracted from an image to few classes in order to find the fraudulent stamp images (mainly governmental, official ones). In the paper seven shape descriptors were explored using the template matching approach. In order to select the best approach their performance was compared with results provided by almost two hundred humans and collected using appropriate inquiry forms.

General shape analysis applied to stamps retrieval from scanned documents

The main purpose of the paper is to present a method of detection, localization and segmentation of stamps (imprints) in the scanned document. It is a very actual topic these days since more and more traditional paper documents are being scanned and stored on digital media. Such digital copy of a stamp may be then used to print a falsified copy of another document. Thus, an electronic version of paper document stored on a hard drive can be taken as a forensic evidence of possible crime. The process of automatic image retrieval on a basis of stamp identification can make the process of crime investigation more efficient. The problem is not trivial since there is no such thing like stamp standard. There are many variations in size, shape, complexity and ink color. It should be remembered that the scanned document may be degraded in quality and the stamp can be placed on relatively complicated background. The algorithm consists of several steps: color segmentation and pixel classification, regular shapes detection, candidates segmentation and verification. The paper includes also some results of selected experiments on real documents having different types of stamps.

Analysis of four polar shape descriptors properties in an exemplary application

The paper presents the results of a comparison of four shape representation techniques that use the transformation of the points from Cartesian to polar coordinates. An exemplary application was selected -- the recognition of characters extracted from polish car license plates. The main goal was the selection of the best polar shape descriptor basing on the experiments performed as well as the discussion on the application of the algorithms based on polar transform in the mentioned problem. The results of the experiment presented in the paper indicate that amongst used methods the Point Distance Histogram (PDH) works the best. The results achieved using the UNL and UNL-Fourier transforms are worse, yet still acceptable.

Pre-processing, extraction and recognition of binary erythrocyte shapes for computer-assisted diagnosis based on MGG images

The paper presents an approach for computer-assisted diagnosis of some diseases (e.g. anaemia, malaria), which are caused by deformations of erythrocyte shapes. In the method firstly the thresholding of the input MGG image is performed, using modified thresholding based on fuzzy measures. Secondly, every cell is localised and extracted. Only the red blood cells are taken for later processing. Using the template matching approach and shape description algorithm every extracted erythrocyte is assigned to one of the twelve classes. Basing on the knowledge about the number of particular unaffected and affected red blood cells the diagnosis can be made. Hence, the possibilities of automatic diagnosis are discussed to stress the potential application of the method.

The west pomeranian university of technology ear database – a tool for testing biometric algorithms

The biometric identification of persons became a very important problem nowadays. Various aspects of criminality, and lately terrorism, forced us to increase the financial outlays allocated for safety measures based on biometrics. Fingerprints and face are the most popular biometric features. However, new modalities are still desirable (moreover, nowadays the usage of multimodal systems becomes very popular). An ear seems to be a very interesting one. The auricle has complex and stable structure and is as distinguishable for various persons as the face. During the development of biometric algorithms their proper and reliable testing is one of the most important aspects. For that purpose a test database can be very helpful. There are many test bases with faces, but ear databases are very rare. Moreover, the existing ones are usually strongly limited. Therefore, in this paper a new ear images database (the West Pomeranian University of Technology Ear Database) is presented for usage in various scientific applications, e.g. for testing biometric algorithms.

Effect of density of measurement points collected from a multibeam echosounder on the accuracy of a digital terrain model

Digital terrain models (DTMs), finding a wide range of applications in the exploration of water areas, are mainly created on the basis of bathymetric data from a multibeam echosounder. The estimation of DTM accuracy dependent on the choice of the survey parameters is difficult due to the lack of reference surface. These authors have developed the methodology of simulation called virtual survey, which enables examining how various parameters of the echosounder, survey and DTM construction algorithms affect the errors of the created models. They are aimed at precise estimation of the model accuracy and the optimization of depth measurement work. The article includes the results of the examination of the effect of parameters determining the density of measurement points on the accuracy of the obtained GRID model. It has been proved that a significant reduction of recorded data density leads to only a slight increase in the modeling error, which makes the bathymetric survey much more cost-effective.

SmartMonitor: an approach to simple, intelligent and affordable visual surveillance system

The paper provides fundamental information about the SmartMonitor --- an innovative surveillance system based on video content analysis. We present a short introduction to the characteristics of the developed system and a brief review of methods commonly applied in surveillance systems nowadays. The main goal of the paper is to describe planned basic system parameters as well as to explain the reason for creating it. SmartMonitor is being currently developed but some experiments have already been performed and their results are provided as well.

An Algorithm for the Automatic Analysis of Characters Located on Car License Plates

The method for the analysis of signs visible on car license plates is described and experimentally evaluated in this paper. The algorithm is applied to the recognition of characters localised on car license plates, used in Poland. Emphasis is put on the especially complex cases — i.e. objects distorted by noise and occlusion — which are the most difficult and challenging tasks of the discussed problem and can be caused, for example, by bad weather conditions. The influence of poor quality images is another source of the problem. The modified UNL transform is applied to solve the problem. This is an algorithm for contour shape representation and recognition, which is not only invariant to affine transformations, but also robust against noise and occlusion. The algorithm is based on the transformation of points belonging to an object that has been extracted from the image from Cartesian to polar coordinates.

A three-dimensional shape description algorithm based on polar-fourier transform for 3D model retrieval

In the paper a new 3D shape representation algorithm is proposed -- the Polar-Fourier 3D Shape Descriptor. Similarly to the Light Field Descriptor, the proposed method is based on rendering several two-dimensional projections of a 3D model, taken from various points of view. However, the proposed descriptor uses the 2D Polar-Fourier transform for obtained projections This enables the new descriptor to be more efficient in the recognition or retrieval of 3D models. In order to evaluate the performance of the algorithm, it was experimentally compared with four other popular approaches -- the Extended Gaussian Image, Shape Distributions, Shape Histogram and Light Field Descriptor -- in the problem of 3D shape retrieval. The achieved results have shown that the new method outperforms the other four explored ones. The presented 3D shape descriptor can be used in representation, recognition and retrieval of 3D models.

Identification of erythrocyte types in greyscale MGG images for computer-assisted diagnosis

In the paper an algorithm for the recognition of erythrocytes is presented and experimentally evaluated. The objects of interest are localised and extracted from digital microscopic images, stained by means of the MGG (May-Grunwald-Giemsa) method in greyscale. The area covering a single red blood cell (RBC) is transformed from Cartesian to polar co-ordinates. Later, the two-dimensional Fourier transform is applied to the resultant image. Finally, the subpart of the spectrum is selected in order to represent an object. This description (Polar-Fourier Greyscale Descriptor) is matched with the templates represented in the same way. The smallest dissimilarity measure indicates the recognised erythrocyte type. When using this approach every RBC is investigated, and basing on the whole knowledge about the number of particular types of erythrocytes present in an image a diagnosis can be made.