Ivo R. Draganov

Response Time Analysis of Text-Based CAPTCHA by Association Rules

The paper introduces and discusses the usability problem of text-based type of CAPTCHA. In particular, two types of text-based CAPTCHA, with text and with numbers, are in the focus. The usability is considered in terms of response time to find a solution for the two aforementioned types of CAPTCHA. To analyze the response time, an experiment is conducted on 230 Internet users, characterized by multiple features, like age, number of years of Internet use, education level, response time in solving text-based CAPTCHA and response time in solving text-number-based CAPTCHA. Then, association rules are extracted from the values of these features, by employing the Apriori algorithm. It determines a new and promising statistical analysis in this context, revealing the dependence of response time to CAPTCHA to the co-occurrence of the feature values and the strength of these dependencies by rule support, confidence and lift analysis.

Compression of CT Images with Branched Inverse Pyramidal Decomposition

In this chapter a new approach is suggested for compression of CT images with branched inverse pyramidal decomposition. A packet of CT images is analyzed and the correlation between each couple inside it is found. Then the packet is split into groups of images with almost even correlation, typically into six or more. One is chosen as a referent being mostly correlated with all of the others. From the rest difference images with the referent are found. After the pyramidal decomposition a packet of spectral coefficients is formed and difference levels which are coded by entropy coder. Scalable high compression is achieved at higher image quality in comparison to that of the JPEG2000 coder. The proposed approach is considered perspective also for compression of MRI images.

Differentiation of the Script Using Adjacent Local Binary Patterns

The paper proposed an algorithm for script discrimination using adjacent local binary patterns (ALBP). In the first stage, each letter is modeled according to its height. The real data are extracted from the probability distribution of the letter heights. Then, the gray scale co-occurrence matrix is computed. It is used as a starting point for the feature extraction. The extracted features are classified according to ALBP. Because of the variety in script characteristics, the statistical analysis shows the differences between scripts. Accordingly, the linear discrimination function is proposed to distinct the scripts. The proposed method is tested on the samples of the printed documents, which include Cyrillic and Glagolitic script. The results of experiments are encouraging.

Multilingual Names Database Searching Enhancement

In this paper some aspects are presented for multilingual names database searching enhancement. They should help to find some balance between the execution time and the relevance of the results obtained. Structured form of the information used, two different structures of the database and a general algorithm for the searching are suggested. As experimental results investigation of the groupings of the most popular Romanized Bulgarian own names produced by the Soundex and Daitch-Mokotoff Soundex (D-M Soundex) algorithms are given. Then conclusion is made for their applicability.

An Adaptive Enhancement of X-Ray Images

Most of the X-ray images are no truly isotropic and its quality varies depending on penetration of X-rays in different anatomical structures and on the technologies of their obtaining. The noise problem arises from the fundamentally statistical nature of photon production. This paper presents an approach for X-ray image enhancement based on contrast limited adaptive histogram equalization (CLAHE), following by morphological processing and noise reduction, based on the Wavelet Packet Decomposition and adaptive threshold of wavelet coefficients in the high frequency sub-bands of the shrinkage decomposition. Implementation results are given to demonstrate the visual quality and to analyze some objective estimation parameters in the perspective of clinical diagnosis.

An Adaptive Approach for Noise Reduction in Sequences of CT Images

CT presents images of cross-sectional slices of the body. The quality of CT images varies depending on penetrating X-rays in a different anatomically structures. Noise in CT is a multi-source problem and arises from the fundamentally statistical nature of photon production. This chapter presents an adaptive approach for noise reduction in sequences of CT images, based on the Wavelet Packet Decomposition and adaptive threshold of wavelet coefficients in the high frequency sub-bands of the shrinkage decomposition. Implementation results are given to demonstrate the visual quality and to analyze some objective estimation parameters such as PSNR, SNR, NRR, and Effectiveness of filtration in the perspective of clinical diagnosis.

Inverse pyramid decomposition of wavelet spectrum for image compression

In this paper, a novel approach is presented for compression of digital images. It consists of finding the wavelet spectrum of an image into certain number of sub-bands after given levels of regular or irregular decomposition. Then each sub-band is decomposed with the inverse pyramid algorithm using some linear orthogonal transform such as DCT. Depending on whether a lossless or lossy compression is desired all or only some of the spectral coefficients from the inverse pyramid are preserved and then entropy coding is applied. Higher compression ratios are achieved at high image quality levels compared to some popular algorithms from the practice.

Exploring the Usability of the Dice CAPTCHA by Advanced Statistical Analysis

This paper introduces a new study of the Dice CAPTCHA usability based on advanced statistical analysis. An experiment is performed on a population of 197 Internet users, characterised by age and Internet experiences, to which the solution to the Dice CAPTCHA is required on a laptop or tablet computer. The response time, which is the solution time to successfully solve the CAPTCHA, together with the number of tries are registered for each user. Then, the collected data are subjected to association rule mining for analysing the dependence of the response time to solve the CAPTCHA in a given number of tries on the co-occurrence of the user’s features. This analysis is very useful to understand the co-occurrence of factors influencing the solution to the CAPTCHA, and accordingly, to realise which CAPTCHA is closer to the “ideal” CAPTCHA.

Self-Organizing Map Classification of the Extremely Low-Frequency Magnetic Field Produced by Typical Tablet Computers

Abstract In this paper, the extremely low frequency magnetic field produced by the tablet computers is explored. The measurement of the tablet computers’ magnetic field is performed by using a measuring geometry previously proposed for the laptop computers. The experiment is conducted on five Android tablet computers. The measured values of the magnetic field are compared to the widely accepted TCO safety standard. Then, the results are classified by the Self-Organizing Map method in order to create different levels of safety or danger concerning the magnetic field to which tablet computer users are exposed. Furthermore, a brief comparison of the obtained magnetic field levels with the ones from typical laptops is performed. At the end, a practical suggestion on how to avoid the high exposure to the low frequency magnetic field emitted by the tablet computers is given.

Classification of Mental Tasks from EEG Signals Using Spectral Analysis, PCA and SVM

Abstract Signals provided by the ElectroEncephaloGraphy (EEG) are widely used in Brain-Computer Interface (BCI) applications. They can be further analyzed and used for thinking activity recognition. In this paper we proposed an algorithm that is able to recognize five mental tasks using 6 channel EEG data. The main idea is to separate the raw EEG signals into several frames and compute their spectrums. Next, a second-order derivative of Gaussian is applied to extract features and an optimum Gaussian kernel parameters grid search is performed with the help of cross-validation. The extracted features are further reduced by Principal Component Analysis. The processed data is utilized to train SVM classifier which is used for mental tasks recognition afterwards. The performance of the algorithm is estimated on publically available dataset. In terms of 5 folds cross-validation we obtained an average of 82.7% recognition rate (accuracy). Additional experiments were conducted using leave-one-out cross-validation where 67.2% correct classification was reported. Comparison to several state-of-the art methods reveals the advantages of the proposed algorithm.