Showing papers on "Face detection published in 1995"

Human Face Detection in Visual Scenes

Abstract: We present a neural network-based face detection system. A retinally connected neural network examines small windows of an image, and decides whether each window contains a face. The system arbitrates between multiple networks to improve performance over a single network. We use a bootstrap algorithm for training, which adds false detections into the training set as training progresses. This eliminates the difficult task of manually selecting non-face training examples, which must be chosen to span the entire space of non-face images. Comparisons with another state-of-the-art face detection system are presented; our system has better performance in terms of detection and false-positive rates.

A unified approach to coding and interpreting face images

Abstract: Face images are difficult to interpret because they are highly variable. Sources of variability include individual appearance, 3D pose, facial expression and lighting. We describe a compact parametrised model of facial appearance which takes into account all these sources of variability. The model represents both shape and grey-level appearance and is created by performing a statistical analysis over a training set of face images. A robust multi-resolution search algorithm is used to fit the model to faces in new images. This allows the main facial features to be located and a set of shape and grey-level appearance parameters to be recovered. A good approximation to a given face can be reconstructed using less than 100 of these parameters. This representation can be used for tasks such as image coding, person identification, pose recovery, gender recognition and expression recognition. The system performs well on all the tasks listed above. >

Automatic face location detection and tracking for model-assisted coding of video teleconferencing sequences at low bit-rates

Abstract: We present a novel and practical way to integrate techniques from computer vision to low bit-rate coding systems for video teleconferencing applications. Our focus is to locate and track the faces of persons in typical head-and-shoulders video sequences, and to exploit the face location information in a ‘classical’ video coding/decoding system. The motivation is to enable the system to selectively encode various image areas and to produce psychologically pleasing coded images where faces are sharper. We refer to this approach as model-assisted coding. We propose a totally automatic, low-complexity algorithm, which robustly performs face detection and tracking. A priori assumptions regarding sequence content are minimal and the algorithm operates accurately even in cases of partial occlusion by moving objects. Face location information is exploited by a low bit-rate 3D subband-based video coder which uses both a novel model-assisted pixel-based motion compensation scheme, as well as model-assisted dynamic bit allocation with object-selective quantization. By transferring a small fraction of the total available bit-rate from the non-facial to the facial area, the coder produces images with better-rendered facial features. The improvement was found to be perceptually significant on video sequences coded at 96 kbps for an input luminance signal in CIF format. The technique is applicable to any video coding scheme that allows for fine-grain quantizer selection (e.g. MPEG, H.261), and can maintain full decoder compatibility.

Face detection by fuzzy pattern matching

Abstract: The paper describes an approach to detect faces whose size and position are unknown in an image with a complex background. The candidates of faces are detected by finding out "face like" regions in the input image using the fuzzy pattern matching method. The perceptually uniform color space is used in our research in order to obtain reliable results. The skin color that is used to detect face like regions, is represented by a model developed by us called skin color distribution function. The skin color regions are then extracted by estimating a measure that describes how well the color of a pixel looks like the skin color for each pixel in the input image. The faces which appear in images are modeled as several 2 dimensional patterns. The face like regions are extracted by a fuzzy pattern matching approach using these face models. The face candidates are then verified by estimating how well the extracted facial features fit a face model which describes the geometrical relations among facial features. >

A methodology for quantitative performance evaluation of detection algorithms

Abstract: We present a methodology for the quantitative performance evaluation of detection algorithms in computer vision A common method is to generate a variety of input images by varying the image parameters and evaluate the performance of the algorithm, as algorithm parameters vary Operating curves that relate the probability of misdetection and false alarm are generated for each parameter setting Such an analysis does not integrate the performance of the numerous operating curves In this paper, we outline a methodology for summarizing many operating curves into a few performance curves This methodology is adapted from the human psychophysics literature and is general to any detection algorithm The central concept is to measure the effect of variables in terms of the equivalent effect of a critical signal variable, which in turn facilitates the determination of the breakdown point of the algorithm We demonstrate the methodology by comparing the performance of two-line detection algorithms

A simple and efficient method for accurate collision detection among deformable polyhedral objects in arbitrary motion

Abstract: We propose an accurate collision detection algorithm for use in virtual reality applications. The algorithm works for three-dimensional graphical environments where multiple objects, represented as polyhedra (boundary representation), are undergoing arbitrary motion (translation and rotation). The algorithm can be used directly for both convex and concave objects and objects can be deformed (non-rigid) during motion. The algorithm works efficiently by first reducing the number of face pairs that need to be checked accurately for interference by first localizing possible collision regions using bounding box and spatial subdivision techniques; face pairs that remain after this pruning stage are then accurately checked for interference. The algorithm is efficient, simple to implement, and does not require any memory intensive auxiliary data structures to be precomputed and updated. Since polyhedral shape representation is one of the most common shape representation schemes, this algorithm should be useful to a wide audience. Performance results are given to show the efficiency of the proposed method.

Face recognition from frontal and profile views

Abstract: This paper presents a unique face recognition sys tem which considers information from both frontal and pro le view images This system represents the rst step toward the development of a face recogni tion solution for the intensity image domain based on a D context In the current system we construct a D face centered model from the two independent images Geometric information is used for view nor malization and at the lowest level the comparison is based on general pattern matching techniques We also discuss the use of geometric information to index the reference database to quickly eliminate impossi ble matches from further consideration The system has been tested using subjects from the FERET program database and has shown excellent results For example we consider the problem of identifying the of the database which is most similar to the target The correct match is included in this list of the time in the system s fully automated mode and of the time in the manually assisted mode The International Workshop on Automatic Face and Gesture Recognition Zurich June

Automatic recognition of human facial expressions

Abstract: The paper presents a new idea for detecting an unknown human face in input imagery and recognizing his/her facial expression represented in the deformation of the two dimensional net, called potential net. The method deals with the facial information, faceness and expressions, as an overall pattern of the net activated by edges in a single input image of face, rather than from changes in the shape of the facial organs or their geometrical relationships. We build models of facial expressions from the deformation patterns in the potential net for face images in the training set of different expressions and then project them into emotion space. Expression of an unknown subject can be recognized from the projection of the net for the image into the emotion space. The potential net is further used to model the common human face. The mosaic method representing energy in the net is used as a template for finding candidates for the face area and the candidates are verified their faceness by projecting them into emotion space in order to select the finalist. Precise location of the face is determined by the histogram analysis of vertical and horizontal projections of edges. >

Learning Human Face Detection in Cluttered Scenes

Abstract: This paper presents an example-based learning approach for locating vertical frontal views of human faces in complex scenes. The technique models the distribution of human face patterns by means of a few view-based “face” and “non-face” prototype clusters. A 2-Value metric is proposed for computing distance features between test patterns and the distribution-based face model during classification. We show empirically that the prototypes we choose for our distribution-based model, and the metric we adopt for computing distance feature vectors, are both critical for the success of our system.

A neural network approach to face/palm recognition

Abstract: This paper proposes a face/palm recognition system based on decision-based neural networks (DBNN). The face recognition system consists of three modules. First, the face detector finds the location of a human face in an image. The eye localizer determines the positions of both eyes in order to generate meaningful feature vectors. The facial region proposed contains eyebrows, eyes, and nose, but excluding mouth. (Eye-glasses will be permissible.) Lastly, the third module is a face recognizer. The DBNN can be effectively applied to all the three modules. It adopts a hierarchical network structures with nonlinear basis functions and a competitive credit-assignment scheme. The paper demonstrates its successful application to face recognition applications on both the public (FERET) and in-house (SCR) databases. In terms of speed, given the extracted features, the training phase for 100-200 persons would take less than one hour on Sparc10. The whole recognition process (including eye localization, feature extraction, and classification using DBNN) may consume only a fraction of a second on Sparc10. Experiments on three different databases all demonstrated high recognition accuracies. A preliminary study also confirms that a similar DBNN recognizer can effectively recognize palms, which could potentially offer a much more reliable biometric feature.

Natural basis functions and topographic memory for face recognition

Abstract: Recent work regarding the statistics of natural images has revealed that the dominant eigenvectors of arbitrary natural images closely approximate various oriented derivative-of-Gaussian functions; these functions have also been shown to provide the best fit to the receptive field profiles of cells in the primate striate cortex. We propose a scheme for expression-invariant face recognition that employs a fixed set of these "natural" basis functions to generate multiscale iconic representations of human faces. Using a fixed set of basis functions obviates the need for recomputing eigenvectors (a step that was necessary in some previous approaches employing principal component analysis (PCA) for recognition) while at the same time retaining the redundancy-reducing properties of PCA. A face is represented by a set of iconic representations automatically extracted from an input image. The description thus obtained is stored in a topographically-organized sparse distributed memory that is based on a model of human long-term memory first proposed by Kanerva. We describe experimental results for an implementation of the method on a pipeline image processor that is capable of achieving near real-time recognition by exploiting the processor's frame-rate convolution capability for indexing purposes. 1 Introduction The problem of object recognition has been a central subject in the field of computer vision. An especially interesting albeit difficult subproblem is that of recognizing human faces. In addition to the difficulties posed by changing viewing conditions, computational methods for face recognition have had to confront the fact that faces are complex non-rigid stimuli that defy easy geometric characterizations and form a dense cluster in the multidimensional space of input images. One of the most important issues in face recognition has therefore been the representation of faces. Early schemes for face recognition utilized geometrical representations; prominent features such as eyes, nose, mouth, and chin were detected and geometrical models of faces given by feature vectors whose dimensions, for instance, denoted the relative positions of the facial features were used for the purposes of recognition [Bledsoe, 1966; Kanade, 1973]. Recently, researchers have reported successful results using photometric representations i.e. representations that are computed directly from the intensity values of the input image. Some prominent examples include face representations based on biologically-motivated Gabor filter "jets" [Buhmann et al., 1990], randomly placed zeroth-order Gaussian kernels [Edelman et a/. This paper explores the use of an iconic representation of human faces that exploits the dimensionality-reducing properties of PCA. However, unlike previous approaches employing …

A mobile robot that recognizes people

Abstract: In order for mobile robots to interact effectively with people they will have to recognize faces. We describe a robot system that finds people, approaches them and then recognizes them. The system uses a variety of techniques: color vision is used to find people; vision and sonar sensors are used to approach them; a template-based pattern recognition algorithm is used to isolate the face; and a neural network is used to recognize the face. All of these processes are controlled using an intelligent robot architecture that sequences and monitors the robot's actions. We present the results of many experimental runs using an actual mobile robot finding and recognizing up to six different people.

Human face detection using silhouettes

Abstract: Face detection is integral to any automatic face recognition system. The goal of this research is to develop a system that performs the task of human face detection automatically in a scene. A system to correctly locate and identify human faces will find several applications, some examples are criminal identification and authentication in secure systems. This work presents a new approach based on principal component analysis. Face silhouettes instead of intensity images are used for this research. It results in reduction in both space and processing time. A set of basis face silhouettes are obtained using principal component analysis. These are then used with a Hough-like technique to detect faces. The results show that the approach is robust, accurate and reasonably fast.

Human‐face search and location in a scene by multi‐pyramid architecture for personal identification

Abstract: To realize fully automated face image recognition, there must be thorough processing from the detection of the face in a scene to recognition. It is very difficult, however, to apply the detection or recognition process based on the shape features as in the conventional method to the face which is a typical soft object. This paper aims at personal identification by the facial image. The face in a scene is sought by coarse-to-fine processing using only the gray-level data, and the result is applied to the recognition. First, the human head is detected from the scene using the multiresolution mosaic. Then the central part of the face is detected from the head region using the mosaic, and the precise position is determined based on the histogram for the eye and the nose region. The search algorithm is applied to 100 personal images derived from the motion image. The detection and location succeeded in 97 percent of the cases, except for the face with eyes shielded by hair, for example. When the result of successful detection is applied to the recognition, the recognition rate of 99 percent is obtained. In this method, a facial image of any size at any position in a scene can be detected. Other features are that the background can be uniform, and the color data are not required, which greatly relaxes the past requirement for the input image.

A subspace method for maximum likelihood target detection

Abstract: We present an unsupervised technique for visual target modeling which is based on density estimation in high-dimensional spaces using an eigenspace decomposition. A computationally efficient and optimal estimator for a multivariate Gaussian distribution is derived. This density estimate is then used to formulate a maximum likelihood estimation framework for visual search and target detection. Our learning technique is applied to the probabilistic visual modeling and subsequent detection of facial features and is shown to be superior to matched filtering.

Decision-based neural network for face recognition system

Abstract: This paper proposes a face recognition system based on decision-based neural networks (DBNN). The DBNN adopts a hierarchical network structure with nonlinear basis functions and a competitive credit-assignment scheme. The face recognition system consists of three modules. First, a face detector finds the location of a human face in an image. Then an eye localizer determines the positions of both eyes to help generate size-normalized, reoriented, and reduced-resolution feature vectors. (The facial region proposed contains eyebrows, eyes, and nose, but excluding mouth. Eye-glasses will be permissible.) The last module is a face recognizer. The DBNN can be effectively applied to all the three modules. The DBNN based face recognizer has yielded very high recognition accuracies based on experiments on the ARPA-FERET and SCR-IM databases. In terms of processing speeds and recognition accuracies, the performance of DBNN is superior to that of multilayer perceptron (MLP). The training phase for 100 persons would take around one hour, while the recognition phase (including eye localization, feature extraction, and classification using DBNN) consumes only a fraction of a second (on Sparc10).

Face location in wavelet-based video compression for high perceptual quality videoconferencing

Abstract: We present a human face location technique based on contour extraction within the framework of a wavelet-based video compression scheme for videoconferencing applications. In addition to an adaptive quantization in which spatial constraints are enforced to preserve perceptually important information at low bit rates, semantic information of the human face is incorporated to design a hybrid compression scheme for videoconferencing, since the face is often the most important part and should be coded with high fidelity. The human face is detected based on contour extraction and feature point analysis. An approximate face mask is then used in the quantization of the decomposed subbands. At the same total bit rate, coarser quantization of the background enables the face region to be quantized finer and coded with a higher quality. Moreover, the resultant larger quantization noise in the background can be suppressed using an edge-preserving enhancement algorithm. Experimental results have shown that the perceptual image quality is greatly improved using the proposed scheme.

Generalization from a Single View in Face Recognition

Abstract: We describe a computational model of face recognition, which generalizes from single views of faces by taking advantage of prior experience with other faces, seen under a wider range of viewing conditions. The model represents face images by vectors of activities of graded overlapping receptive fields (RFs). It relies on high spatial frequency information to estimate the viewing conditions, which are then used to normalize (via a transformation specific for faces), and identify, the low spatial frequency representation of the input. The class-specific transformation approach allows the model to replicate a series of psychophysical findings on face recognition, and constitutes an advance over current face recognition methods, which are incapable of generalization from a single example.

Use of face robot for human-computer communication

Abstract: We are attempting to introduce a 3-dimensional, realistic human-like face robot to human-computer communication modality. The face robot can recognize human facial expressions as well as produce more realistic facial expressions. We propose a new concept of "Active Human Interface"; and as the first step, we investigate the performance of facial expression recognition by neural network (NN) and the expressionability of facial messages on face robot. We find that the NN recognition of facial expressions and face robot's performance in generating facial expressions are of almost same level as that in humans. This implies a high potential in the use of face robot for human-computer communication media.

Towards an automatic human face localization system

Abstract: This paper describes a method to detect and locate human faces in an image given no prior information about the size, orientation, and viewpoint of the faces in the image. This method uses a family of Gaussian derivative filters to search and extract human facial features from the image and then group them together into a set of partial faces using their geometric relationship. A belief network is then constructed for each possible face candidate and the belief values updated by evidences propagating through the network. Different instances of detected faces are then compared using their belief values and improbable face candidates discarded. The algorithm is tested on different instances of faces with varying sizes, orientation and viewpoint and the results indicate a 91% success rate in detection under viewpoint variation.

Parallel algorithms for real-time colliding face detection

Abstract: We propose parallel algorithms for detecting collisions among 3D objects in real-time. First, a basic algorithm of serial version is described. It can detect potential collisions among multiple objects with arbitrary motion (translation and rotation) in 3D space. The algorithm can be used without modification for both convex and concave objects represented as polyhedra. This algorithm is efficient, simple to implement, and does not require any memory intensive auxiliary data structure to be precomputed and updated. Then, two parallel algorithms are proposed for MIMD multi-processors having a shared-memory; one uses a static and the other uses a dynamic method for proper load balancing. Experimental results demonstrate the performance of the proposed collision detection methods.

Preprocessing images of faces: correlations with human perceptions of distinctiveness and familiarity

Abstract: The aim of the paper work is to further our understanding of how humans process and recognise faces. The authors do this by proceeding in parallel with testing subjects and building computer models. If a model reflects the way that humans process face images, it ought to fail (in the same way) to find the same faces easy or difficult. One characteristic of human recognition is that of distinctiveness: some faces are never forgotten, others easily lost in a crowd. The present paper describes the use of various forms of image processing to see whether they correlate with human perceptions of distinctiveness, memorability and familiarity.

Frontal-view face detection

Abstract: This paper presents a symmetry measurement based on the correlation coefficient. This symmetry measure-ment is used to locate the center line of faces, and afterward, to decide whether the face view is frontal or not. A483-face image database obtained from the U. S. Army was used to test the algorithm. Though the performance ofthe algorithm is limeted to 87%, this is due to the wide range of variations present in the database used to test ouralgorithm. Under more constrain conditions, such as uniform illumination, this technique can be a powerful toolin facial feature extraction. In regards its computational requirements, though this algorithm is very expensive,three independent optimizations are presented; two of which are successfully implemented, and tested.Keywords: symmertry, symmetry measurement, face detection, frontal view face detection. 1 INTRODUCTION. This paper presents an algorithm based on symmetry measurements, useful to extract information about sym-metric objects from images. The proposed technique was developed and tested in the context of face recognition,but it might find applications in other areas of computer vision in which symmetry is involved. The motivationfor this work is the high amount of symmetry present in frontal view faces; the claim is that this information canbe useful in the estimation of face orientation, as well as in the extraction of feature points.In a face recognition system based on template matching, estimating the orientation of the probe faces helps indiscarding templates that do not need to be searched reducing the computational requirements and the executiontime.1 On the other hand, this algorithm as a preprocessing step, provides information usually assumed as inputdata in facial feature extraction techniques. In a more general context, symmetry measurements combined withsome knowledge can be used to verify the presence of (symmetrical) objects in a given image.Much work has been previously done in this area; from the development of symmetry operators24 for detectionof interesting points, to fast algorithms for the location of axis of symmetry on images.5 However, the problemhas been stated as the location of the regions with the largest amount of symmetry to guide the feature extractionassuming their presence. In the context of face recognition, we use a symmetry measurement to decide if theprobe image is a frontal view; for each case, we also provide an estimation of the tilt angle, and the location ofthe center line.

Appropriate-scale local centers: a foundation for parts-based recognition

Abstract: An image representation in terms of local centers is developed and motivated in the context of natural object recognition. Local centers are visually compact regions that have significant internal-external contrast in some measurement. Existing computational models of the concept are compared, and a particular model, the appropriate-scale ridge, is developed. In this model, local centers are defined as smoothed image extrema that are also extrema, with respect to scale, in the second spatial derivatives. The usefulness of the model is analyzed via 1D scale-space behavior and demonstrated on 2D natural images. The basic concept of a local center is also extended to color and shading data. The stability of the shading model is demonstrated on images of a moving face under varying illumination and tested via a face detection task.

Learning a distribution-based face model for human face detection

Abstract: We present a distribution-based modeling cum example-based learning approach for detecting human faces in cluttered scenes. The distribution-based model captures complex variations in human face patterns that cannot be adequately described by classical pictorial template-based matching techniques or geometric model-based pattern recognition schemes. We also show how explicitly modeling the distribution of certain "facelike" nonface patterns can help improve classification results.

Segmentation, grouping and feature detection for face image analysis

Abstract: Automatic face and facial feature detection by a computer is a computationally challenging problem. There are many shape patterns for eyes, nose and mouth, hair, beards/mustaches, and the overall geometry of the face frame. The texture and color of hair and skin also vary greatly. One source of complexity in analyzing face images is due to the different appearances resulting from different orientations. From the literature and the authors' own experience, it is apparent that the various face orientations, eyeglasses, and hair, are among the main factors affecting the complexity of the face image analysis problems. This paper discusses a system that analyzes face images under a wide range of orientations, like those in the FERET database. Unlike previous systems that were designed specifically for either frontal or side-view images but not both, this takes a unified approach and treats frontal and side-view images with the same common sequence of analysis, with view-specific analysis only at the final stage, amounting to less than 10% of the whole task. It also detects the presence of eyeglasses under all views.