Projection based method for segmentation of human face and its evaluation

We detect facial features and then circumscribe each facial feature with the smallest rectangle possible by using vertical and horizontal gray value projections of pixels. The result is evaluated with respect to the manually located enclosing rectangle on the images of a publicly available database.     

A spectral independent approach for physiological and geometric based face recognition in the visible,middle-wave and long-wave infrared bands

The difficulty of face recognition (FR) systems to operate efficiently in diverse operational environments, e.g. day and night time, is aided by employing sensors covering different spectral bands (i.e. visible and infrared). Biometric practitioners have identified a framework of band-specific algorithms, which can contribute to both assessment and intervention. While these motions are proven to achieve improvement of identification performance, they traditionally result in solutions that typically fail to work efficiently across multiple spectrums. In this work, we designed and developed an efficient, fully automated, direct matching-based FR approach, that is designed to operate efficiently when face data is captured using either visible or passive infrared (IR) sensors. Thus, it can be applied in both daytime and nighttime environments. First, input face images are geometrically normalized using our pre-processing pipeline prior to feature-extraction. Then, face-based features including wrinkles, veins, as well as edges of facial characteristics, are detected and extracted for each operational band (visible, MWIR, and LWIR). Finally, global and local face-based matching is applied, before fusion is performed at the score level. Our approach achieves a rank-1 identification rate of at least 99.43%, regardless of the spectrum of operation. This suggests that our approach results in better performance than other tested standard commercial and academic face-based matchers, on all spectral bands used.     

Zeros of equivariant vector fields: Algorithms for an invariant approach

We present a symbolic algorithm to solve for the zeros of a polynomial vector field equivariant with respect to a finite subgroup of O (n). We prove that the module of equivariant. polynomial maps for a finite matrix group is Cohen-Macaulay and give an algorithm to compute a fundamental basis. Equivariant normal forms are easily computed from this basis. We use this basis to transform the problem of finding the zeros of an equivariant map to the problem of finding zeros of a set of invariant polynomials. Solving for the values of fundamental polynomial invariants at the zeros effectively reduces each group orbit of solutions to a single point. Our emphasis is on a computationally effective algorithm and we present our techniques applied to two examples.     

Dissimilarity criteria and their comparison for quantitative evaluation of image segmentation: application to human retina vessels

The quantitative evaluation of image segmentation is an important and difficult task that is required for making a decision on the choice of a segmentation method and for the optimal tuning of its parameter values. To perform this quantitative evaluation, dissimilarity criteria are relevant with respect to the human visual perception, contrary to metrics that have been shown to be visually not adapted. This article proposes to compare eleven dissimilarity criteria together. The field of retina vessels image segmentation is taken as an application issue to emphasize the comparison of five specific image segmentation methods, with regard to their degrees of consistency and discriminancy. The DRIVE and STARE databases of retina images are employed and the manual/visual segmentations are used as a reference and as a control method. The so-called \(\epsilon \) criterion gives results in agreement with perceptually based criterions for achieving the quantitative comparison.     

基于图像分块的LDA人脸识别

设计了一种基于图像分块的LDA(linear discriminant analysis)人脸识别方法,该方法从模式的原始数字图像出发,先对图像矩阵进行分块,然后对分块子图像进行LDA特征提取,从而得到能代替原始模式的低维新模式,最后再用最小距离分类器进行分类.该方法克服了传统LDA方法的缺点,其优点是能有效地提取图像的局部特征.实验表明:该方法在识别性能上优于Fisheffaces方法.     

Aligned texture map creation for pose invariant face recognition

In last years, Face recognition based on 3D techniques is an emergent technology which has demonstrated better results than conventional 2D approaches. Using texture (180° multi-view image) and depth maps is supposed to increase the robustness towards the two main challenges in Face Recognition: Pose and illumination. Nevertheless, 3D data should be acquired under highly controlled conditions and in most cases depends on the collaboration of the subject to be recognized. Thus, in applications such as surveillance or control access points, this kind of 3D data may not be available during the recognition process. This leads to a new paradigm using some mixed 2D-3D face recognition systems where 3D data is used in the training but either 2D or 3D information can be used in the recognition depending on the scenario. Following this concept, where only part of the information (partial concept) is used in the recognition, a novel method is presented in this work. This has been called Partial Principal Component Analysis (P²CA) since they fuse the Partial concept with the fundamentals of the well known PCA algorithm. This strategy has been proven to be very robust in pose variation scenarios showing that the 3D training process retains all the spatial information of the face while the 2D picture effectively recovers the face information from the available data. Furthermore, in this work, a novel approach for the automatic creation of 180° aligned cylindrical projected face images using nine different views is presented. These face images are created by using a cylindrical approximation for the real object surface. The alignment is done by applying first a global 2D affine transformation of the image, and afterward a local transformation of the desired face features using a triangle mesh. This local alignment allows a closer look to the feature properties and not the differences. Finally, these aligned face images are used for training a pose invariant face recognition approach (P²CA).     

Algorithms for compact letter displays: Comparison and evaluation

Multiple pairwise comparisons are one of the most frequent tasks in applied statistics. In this context, letter displays may be used for a compact presentation of results of multiple comparisons. A heuristic previously proposed for this task is compared with two new algorithmic approaches. The latter rely on the equivalence of computing compact letter displays to computing clique covers in graphs, a problem that is well-studied in theoretical computer science. A thorough discussion of the three approaches aims to give a comparison of the algorithms’ advantages and disadvantages. The three algorithms are compared in a series of experiments on simulated and real data, e.g., using data from wheat and triticale yield trials.     

Algorithms for unimodal segmentation with applications to unimodality detection

We study the problem of segmenting a sequence into k pieces so that the resulting segmentation satisfies monotonicity or unimodality constraints. Unimodal functions can be used to model phenomena in which a measured variable first increases to a certain level and then decreases. We combine a well-known unimodal regression algorithm with a simple dynamic-programming approach to obtain an optimal quadratic-time algorithm for the problem of unimodal k-segmentation. In addition, we describe a more efficient greedy-merging heuristic that is experimentally shown to give solutions very close to the optimal. As a concrete application of our algorithms, we describe methods for testing if a sequence behaves unimodally or not. The methods include segmentation error comparisons, permutation testing, and a BIC-based scoring scheme. Our experimental evaluation shows that our algorithms and the proposed unimodality tests give very intuitive results, for both real-valued and binary data. Niina Haiminen received the M.Sc. degree from the University of Helsinki in 2004. She is currently a Graduate Student at the Department of Computer Science of University of Helsinki, and a Researcher at the Basic Research Unit of Helsinki Institute for Information Technology. Her research interests include algorithms, bioinformatics, and data mining. Aristides Gionis received the Ph.D. degree from Stanford University in 2003, and he is currently a Senior Researcher at the Basic Research Unit of Helsinki Institute for Information Technology. His research experience includes summer internship positions at Bell Labs, AT&T Labs, and Microsoft Research. His research areas are data mining, algorithms, and databases. Kari Laasonen received the M.Sc. degree in Theoretical Physics in 1995 from the University of Helsinki. He is currently a Graduate Student in Computer Science at the University of Helsinki and a Researcher at the Basic Research Unit of Helsinki Institute for Information Technology. His research is focused on algorithms and data analysis methods for pervasive computing.     

人脸检测及分割方法研究

人脸检测和分割是进行人脸分析和识别的重要组成部分,其目的是从复杂背景图像中检测出人脸的位置,并把人脸分割出来。虽然人们可以毫不困难地在复杂背景中检测和识别出人脸,但是要建立一套完全自动化的人脸和识别系统却是非常困难的。近年来随着计算机技术和图像处理方法的不断发展,出现了各种不同的人脸检测方法。应用了一种基于YIQ颜色模型的人脸检测方法,并且利用惯量最小化原理进行人脸图像的姿态调整,然后进行分割提取。试验结果表明,该方法能较好地从复杂背景中检测出人脸区域并能较完整地分割出人脸。     

一种基于人脸分割的PCA和SVM人脸识别方法

基于PCA和SVM提出了一种新的人脸分割法,将双眼、额头、鼻子、嘴等可以明显表征人脸的六类特征提取出来,舍弃双颊以及耳朵等特征量少的部位。融合上述人脸的特征识别结果,运用支持向量机的方法进行分类识别,实验结果表明,文章所提出的PCA与SVM融合的新的人脸分割方法能有效地对人脸进行分类,极大地提升识别率。     

Human segmentation of infrared image for mobile robot search

In the search robotics field, human target segmentation method plays a basic preprocessing step in the visual guidance. However, with the wide application of the infrared sensor on robot vision, traditional segmentation methods are facing more challenges of low-contrast, overlapping and blurring targets, and complex background. This paper introduces an infrared human segmentation approach that integrates the improved pulse coupled neural network (PCNN), the curvature gravity gradient tensor (CGGT) and the mathematical morphology to address these above problems. This approach starts with an improved PCNN segmentation model. Local dynamic synapse weights are designed to enhance the synchronous pulsing ability of the improved PCNN model with similar inputs, and a reformed threshold is conducted to guide the process of segmentation. Moreover, eigenvalues of CGGT are guaranteed in this model as linking coefficients, in order to capture the edges and details of human target more exactly in segmentation. Lastly, the segmentation result is repaired by morphology operators, to ensure the integrity of the target region and the independent noise removal. Experiments on 200 real infrared images captured by the mobile robot CQSearcher I, demonstrate that our method is superior over the other classic segmentation methods in both the subjective visual performance and the objective indicators of misclassification error and f-measure.     

Feature map sharing hypercolumn model for shift invariant face recognition

In this article, we propose a shift-invariant pattern recognition mechanism using a feature-sharing hypercolumn model (FSHCM). To improve the recognition rate and to reduce the memory requirements of the hypercolumn model (HCM), a shared map is constructed to replace a set of local neighborhood maps in the feature extraction and feature integration layers. The shared maps increase the ability of the network to deal with translation and distortion variations in the input image. The proposed face recognition system employed a FSHCM neural network to perform feature extraction and use a linear support vector machine for a recognition task. The effectiveness of the proposed approach is verified by measuring the recognition accuracy using the misaligned ORL face database.     

A composite spatio-temporal modeling approach for age invariant face recognition

In this research we propose a novel method of face recognition based on texture and shape information. Age invariant face recognition enables matching of an image obtained at a given point in time against an image of the same individual obtained at an earlier point in time and thus has important applications, notably in law enforcement. We investigate various types of models built on different levels of data granularity. At the global level a model is built on training data that encompasses the entire set of available individuals, whereas at the local level, data from homogeneous sub-populations is used and finally at the individual level a personalized model is built for each individual. We narrow down the search space by dividing the whole database into subspaces for improving recognition time. We use a two-phased process for age invariant face recognition. In the first phase we identify the correct subspace by using a probabilistic method, and in the second phase we find the probe image within that subspace. Finally, we use a decision tree approach to combine models built from shape and texture features. Our empirical results show that the local and personalized models perform best when rated on both Rank-1 accuracy and recognition time.     

Illumination invariant extraction for face recognition using neighboring wavelet coefficients

The features of a face can change drastically as the illumination changes. In contrast to pose position and expression, illumination changes present a much greater challenge to face recognition. In this paper, we propose a novel wavelet based approach that considers the correlation of neighboring wavelet coefficients to extract an illumination invariant. This invariant represents the key facial structure needed for face recognition. Our method has better edge preserving ability in low frequency illumination fields and better useful information saving ability in high frequency fields using wavelet based NeighShrink denoise techniques. This method proposes different process approaches for training images and testing images since these images always have different illuminations. More importantly, by having different processes, a simple processing algorithm with low time complexity can be applied to the testing image. This leads to an easy application to real face recognition systems. Experimental results on Yale face database B and CMU PIE Face Database show that excellent recognition rates can be achieved by the proposed method.     

Algorithms for defining visual regions-of-interest: comparison witheye fixations

Many machine vision applications, such as compression, pictorial database querying, and image understanding, often need to analyze in detail only a representative subset of the image, which may be arranged into sequences of loci called regions-of-interest (ROIs). We have investigated and developed a methodology that serves to automatically identify such a subset of aROIs (algorithmically detected ROIs) using different image processing algorithms (IPAs), and appropriate clustering procedures. In human perception, an internal representation directs top-down, context-dependent sequences of eye movements to fixate on similar sequences of hROIs (human identified ROIs). In the paper, we introduce our methodology and we compare aROIs with hROIs as a criterion for evaluating and selecting bottom-up, context-free algorithms. An application is finally discussed     

OptiFuzz: a robust illumination invariant face recognition system and its implementation

Vision-based human face detection and recognition are widely used and have been shown to be effective in normal illumination conditions. Under severe illumination conditions, however, it is very challenging. In this paper, we address the effect of illumination on the face detection and the face recognition problem by introducing a novel illumination invariant method, called OptiFuzz. It is an optimized fuzzy-based illumination invariant method to solve the effect of illumination for photometric-based human face recognition. The rule of the Fuzzy Inference System is optimized by using a genetic algorithm. The Fuzzy’s output controls an illumination invariant model that is extended from Land’s reflectance model. We test our method by using Yale B Extended and CAS-PEAL face databases to represent the offline experiments, and several videos are recorded at our campus to represent the online indoor and outdoor experiments. Viola–Jones face detector and mutual subspace method are employed to handle the online face detection and face recognition experiments. Based on the experimental results, we can show that our algorithm outperforms the existing and the state-of-the-art methods in recognizing a specific person under variable lighting conditions with a significantly improved computation time. Other than that, using illumination invariant images is also effective in improving the face detection performance.     

A multi-classifier approach to face image segmentation for travel documents

The adoption of face images in machine readable travel documents requires some quality constraints to be fulfilled (e.g., no flash reflections on skin or hair across eyes), as specified in the ISO/IEC 19794-5 standard. Automatically evaluating the compliance of a face image to such requirements needs a precise knowledge of the image structure, intended as the partitioning of the image into its main components (face, hair, clothes and background regions). In this paper a multi-classifier system based on color and texture information is proposed for face image segmentation. Extensive experiments carried out both on the segmentation algorithm and on its application to ISO/IEC 19794-5 standard compliance verification are reported and discussed. The results obtained are encouraging and confirm that: (i) the robustness of the proposed segmentation approach to deal with difficult image characteristics (e.g., uneven illumination or varied background) is satisfactory and (ii) the knowledge deriving from image segmentation is very useful for ISO/IEC 19794-5 standard compliance verification.