Paweł Forczmański

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.

Recognition of Occluded Faces Based on Multi-subspace Classification

In the paper we investigate a problem of face recognition in uncontrolled environment – distorted by occlusion, shadows and other local modifications. Such problems are very common for real-world conditions, thus the presented algorithm allows to eliminate them. It is based on dimensionality reduction approach (two-dimensional Karhunen-Loeve Transform) and distance-based classification. We use simple transformations involving face normalization and individual facial regions extraction as a pre-processing. Then, we perform independent recognition of extracted facial regions and combine the results in order to make a final classification. The results of experiments conducted on images taken from 9 publicly available datasets show that a quite simple algorithm is capable of successful recognition without high computing power requirements, as opposite to more sophisticated methods presented in the literature. As it was proved, the presented approach gives significantly better efficiency than a whole image-based recognition.

SmartMonitor--an intelligent security system for the protection of individuals and small properties with the possibility of home automation.

“SmartMonitor” is an intelligent security system based on image analysis that combines the advantages of alarm, video surveillance and home automation systems. The system is a complete solution that automatically reacts to every learned situation in a pre-specified way and has various applications, e.g., home and surrounding protection against unauthorized intrusion, crime detection or supervision over ill persons. The software is based on well-known and proven methods and algorithms for visual content analysis (VCA) that were appropriately modified and adopted to fit specific needs and create a video processing model which consists of foreground region detection and localization, candidate object extraction, object classification and tracking. In this paper, the “SmartMonitor” system is presented along with its architecture, employed methods and algorithms, and object analysis approach. Some experimental results on system operation are also provided. In the paper, focus is put on one of the aforementioned functionalities of the system, namely supervision over ill persons.

Face recognition using two-dimensional CCA and PLS

This paper presents the implementation of the method of twodimensional Canonical Correlation Analysis (CCA) and two-dimensional Partial Least Squares (PLS) applied to image matching. Both methods are based on representing the image as the sets of its rows and columns and implementation of CCA using these sets (hence we named the methods as CCArc and PLSrc). CCArc and PLSrc feature simple implementation and lesser complexity than other known approaches. In applications to biometrics, CCArc and PLSrc are suitable to solving the problems when dimension of images (dimension of feature space) is greater than the number of images, i.e., Small Sample Size (SSS) problem. This paper demonstrates high efficiency of CCArc and PLSrc for a number of computer experiments, using benchmark image databases.

Simple and Robust Facial Portraits Recognition under Variable Lighting Conditions Based on Two-Dimensional Orthogonal Transformations

The paper addresses the problem of face recognition for images registered in variable lighting, which is common for real-world conditions. Presented algorithm is based on orthogonal transformation preceded by simple transformations comprising of equalization of brightness gradients, removal of spatial low frequency spectral components and fusion of spectral features depending on average pixels intensity. Two types of transformations: 2DDCT (two-dimensional Discrete Cosine Transform) and 2DKLT (two-dimensional Karhunen-Loeve Transform) were investigated in order to find the most optimal algorithm setup. The results of experiments conducted on Yale B and Yale B+ datasets show that a quite simple algorithm is capable of successful recognition without high computing power demand, as opposite to several more sophisticated methods presented recently.

Near-Lossless PCA-Based Compression of Seabed Surface with Prediction

The paper presents a compression method based on Principal Component Analysis applied to reduce the volume of data in seabed Digital Terrain Model. Such data have to be processed in a manner very different from typical digital images because of practical aspects of analysed problem. Hence, the developed algorithm features a variable compression ratio and a possibility to control a maximal reconstruction error. The main objective is to build an orthogonal base and find a number of PCA coefficients representing analysed surface with an acceptable reconstruction accuracy. We present two variants of processing: an iterative compression approach and an approach predicting a number of coefficients before compression starts. It yields much lower computational demand and is faster. The later algorithm employs several statistical measures of an input surface describing its complexity at the prediction stage. Employed, simple classifier based on Classification and Regression Tree do not introduce high additional time overhead. Performed experiments on real data showed high compression ratios, better than for typical DCT-based methods. The possible application of developed method is modern data management system employed in maritime industry.

Comparative analysis of benchmark datasets for face recognition algorithms verification

The paper presents a problem of recognition of facial portraits in the aspect of benchmark database quality. The aim of the work presented here was to analyse the potential of datasets published over the Internet and the predicted applicability of such data for the task of face recognition performance verification. We gathered 41 datasets created and published by various academic and commercial bodies. In the paper we focus on both pure data characteristics, including the number of images, their spatial resolution, quality, content and usability, as well as more high-level properties, e.g. face orientation, expression, background, lighting, and attributes like hats, glasses and beards. We have chosen several datasets on which we performed more detailed experiments related to face recognition. We employed several database preparation algorithms (cross-validation based on different schemes) to make the results as much objective as possible. Here, Principal Component Analysis was employed, as a standard tool for dimensionality reduction. The classification was performed using simple Euclidean metrics. Performed experiments showed a true potential of selected databases.

Improving the Recognition of Occluded Faces by Means of Two-dimensional Orthogonal Projection into Local Subspaces

The paper presents a problem of reducing the influence of natural occlusion on face recognition accuracy. It is based on transformation (two-dimensional Karhunen-Loeve Transform) of face parts into local subspaces calculated by means of two-dimensional Principal Component Analysis and two-dimensional Linear Discriminant Analysis. We use a sequence of operations consisting of face scale and orientation normalization and individual facial regions extraction. Independent recognitions are performed on extracted facial regions and their results are combined in order to perform a final classification. The experiments on images taken from publicly available datasets show that such a simple algorithm is able to successfully recognize faces without high computational overhead, in contrast to more sophisticated methods presented recently. In comparison to typical, whole-face-based approach, developed method gives significantly better accuracy.

Application of foreground object patterns analysis for event detection in an innovative video surveillance system

SmartMonitor is an innovative surveillance system based on video content analysis. It is a modular solution that can work in several predefined scenarios mainly concerned with home/surrounding protection against unauthorized intrusion, supervision over ill person and crime detection. Each scenario is associated with several actions and conditions, which imply the utilization of algorithms with various input parameters. In this paper, focus is put on the analysis of foreground object patterns for the purposes of event recognition, as well as the experimental investigation of selected methods and algorithms which were developed and employed for the SmartMonitor system prototype. The prototype performs three main tasks: detection and localization of foreground regions using adaptive background modelling based on Gaussian Mixture Models, candidate objects extraction and classification using Haar and HOG descriptors, and tracking using Mean-Shift algorithm. The main goal of the work described here is to match system parameters with each scenario to provide the highest effectiveness and to decrease the number of false alarms.

Similarity Estimation of Textile Materials Based on Image Quality Assessment Methods

In this paper some experimental results obtained by the application of various image quality assessment methods for the estimation of similarity of textile materials are presented. Such approach is considered as a part of an artificial intelligence based system developed for the recognition of clothing styles based on multi-dimensional analysis of descriptors and features.