Efficient Object Recognition and Segmentation with Enhanced Skin Color Manipulation Technique

KaaShiv InfoTech, Number 1 Inplant Training Experts in Chennai.


Efficient Object Recognition and Segmentation with Enhanced Skin Color Manipulation Technique


Advancements in technology lead to many new inventions. Among them Face Detection process plays a key role in imaging processes. This process analysis for skin color in an image and returns the face location.Earlier personal computers or digital cameras were used to detect faces. But to its limitation in resources, Embedded Smart Cameras or Mobile Devices with built in cameras are preferred nowadays due to its general purpose processors. Our proposed approach scans the targeted image fully and identifies the face through a multi-layered algorithmic process, P-FAD.The P-FAD process is a hierarchical detection scheme which first performs skin detection coarsely and then contour point detection, dynamic group, region merge & filter.Finally modified V-J detector provides the refined results.This methodology helps in reduction of pixel manipulation, computation and storage cost at the same without losing the performance factors.This proposed methodology is tested with video and image detection and the results prove to be efficient.

Segmentation with Enhanced Skin Color Manipulation Technique


Advancements in technology lead to many new inventions. Among them Face Detection process plays a key role in imaging processes. This process analysis for skin color in an image and returns the face location. Earlier personal computers or digital cameras were used to detect faces. But to its limitation in resources, Embedded Smart Cameras or Mobile Devices with built in cameras are preferred nowadays due to its general purpose processors. Our proposed approach scans the targeted image fully and identifies the face through a multi-layered algorithmic process, P-FAD. The P-FAD process is a hierarchical detection scheme which first performs skin detection coarsely and then contour point detection, dynamic group, region merge & filter. Finally modified V-J detector provides the refined results. This methodology helps in reduction of pixel manipulation, computation and storage cost at the same without losing the performance factors. This proposed methodology is tested with video and image detection and the results prove to be efficient.


kaashiv infotech


kaashiv infotech

Related URL’s for reference:

[1]An Efficient Skin Illumination Compensation Model for Efficient Face Detection:


Face detection (human) plays an important role in applications such as human computer interface, face recognition video surveillance and face image database management. In the human face detection applications, face(s) most frequently form an inconsequential part of the images. Consequently, preliminary segmentation of images into regions that contain "non-face" objects and regions that may contain "face" candidates can greatly accelerate the process of human face detection. Most existingface detection approaches have assumptions, which make them applicable only under some specific conditions. Existing techniques for face detection in color images are plagued by poor performance in the presence of scale variation, variation in illumination, variation in skin colors, complex backgrounds etc. In this research work we have made a humble attempt to propose an algorithm for face detection in color images in the presence of varying lighting conditions, for varied skin colors as well as with complex backgrounds. Based on a novel tangible skin component extraction modus operandi and detection of the valid face candidates, our method detects skin regions over the entire image and engenders face candidates based on the signatures of the detected skin patches. The algorithm constructs the boundary for each face candidate. Experimental results demonstrate successful face detection over a wide range of facial variations in color, position, scale, varying lighting conditions, orientation, 3D pose, and expression in images from the database which is enriched with several photo collections (both indoors and outdoors with all the above mentioned extreme cases taken into consideration). In view of the capability to handle high degree of variability, that our approach can handle we articulate that our approach outperforms all the published popular methods.

[2] Analysis of Detection and Track on Partially Occluded Face:


As a new field of information technology, biometrics recognition consists of the recognition of face, iris, retina, pronunciation etc. In recent years, Static face detection algorithms concerning the categories of detection technology have been raised, but these algorithms put their focus on the detection of clear face. The research of detection on partially occluded face has not yet been conducted. Under this circumstance, the thesis gives a brand new static detection and dynamic track algorithm on partially occluded face. This new algorithm is that the static partial face is detected by the YCbCr, the dynamic image sequences use a logic "and" operation to track the partial face, and then the static partial face and dynamic face are fitted, if the fitting value is smaller than the pre-value, the static partial face is considered as a non-face; otherwise the partial is a face. By analyzing the 50 graphs including 160faces (60 partial faces and 100 clear faces), it has been found that the detection accuracy of the partial faces is 62% and the clear faces detection accuracy is 98%. This new algorithm can provide important reference to partial face detection. The algorithm still needs to be verified and improved further.

[3] Face detection based on half face-template:


Face detection in the image is an important research branch of face recognition. For the purpose of detecting the faces in images efficiently, a face detection method based on half face-template is proposed. According to the character of density of the feature organs such as eye, ear, nose, mouth, part of cheek in the face images, the obverse average full face-template is constructed. And the obverse average half face-template is constructed directly based on density symmetry of face-template. The face detection experiments in the images were carried on using the method of template matching and determine the position of the face in the image according to comparability. The theory analysis shows that the average face-template can reduce the chanciness of local density of the template effectively and the half face-template can reduce the symmetry redundancy of density in the face-template and increase the speed of face detection. The experimental result indicates that the half face-template can adapt to side face images in a large angle, which improves the correctness of side face detection substantially.

[4] Multi-feature driver face detection based on area coincidence degree and prior knowledge


Exact and fast driver face detection is a key for recognizing whether drivers are fatigue or not by detecting facial organs while driving using machine vision technology. Aiming at the limitation of driverface detection algorithm based on single feature in detection precision and reliability, a novel fusion algorithm of driver face detection is proposed. Firstly, an improved face detection algorithm based on Haar-like feature is used to detect the possibly existing initial face region in the whole image, then the initial face region detected is extended properly and a face detection algorithm based on skin color feature in rgb space is used to locate the face region again in the extended area, finally, fusion detectionof driver face region is achieved by the defined area coincidence degree and geometric prior knowledge of human face. Experiments carried out in various complicated road environments show the algorithm proposed is of strong robustness on lighting changes, driver head rotation, and having glasses, etc., while at detection precision and reliability, it offers a noticeable enhancement compared with the single feature based algorithms.

[5] Regularized Transfer Boosting for Face Detection Across Spectrum:


This letter addresses the problem of face detection in multispectral illuminations. Face detection in visible images has been well addressed based on the large scale training samples. For the recently emerging multispectral face biometrics, however, the face data is scarce and expensive to collect, and it is usually short of face samples to train an accurate face detector. In this letter, we propose to tackle the issue of multispectral face detection by combining existing large scale visible face images and a few multispectral face images. We cast the problem of face detection across spectrum into the transfer learning framework and try to learn the robust multispectral face detector by exploring relevant knowledge from visible data domain. Specifically, a novel Regularized Transfer Boosting algorithm named R-TrBoost is proposed, with features of weighted loss objective and manifold regularization. Experiments are performed with face images of two spectrums, 850 nm and 365 nm, and the results show significant improvement on multispectral face detection using the proposed algorithm.

[6] A video-based face detection and recognition system using cascade face verification modules:


Face detection (human) plays an important role in applications such as human computer interface, face recognition video surveillance and face image database management. In the human face detection applications, face(s) most frequently form an inconsequential part of the images. Consequently, preliminary segmentation of images into regions that contain "non-face" objects and regions that may contain "face" candidates can greatly accelerate the process of human face detection. Most existingface detection approaches have assumptions, which make them applicable only under some specific conditions. Existing techniques for face detection in color images are plagued by poor performance in the presence of scale variation, variation in illumination, variation in skin colors, complex backgrounds etc. In this research work we have made a humble attempt to propose an algorithm for face detection in color images in the presence of varying lighting conditions, for varied skin colors as well as with complex backgrounds. Based on a novel tangible skin component extraction modus operandi and detection of the valid face candidates, our method detects skin regions over the entire image and engenders face candidates based on the signatures of the detected skin patches. The algorithm constructs the boundary for each face candidate. Experimental results demonstrate successful face detection over a wide range of facial variations in color, position, scale, varying lighting conditions, orientation, 3D pose, and expression in images from the database which is enriched with several photo collections (both indoors and outdoors with all the above mentioned extreme cases taken into consideration). In view of the capability to handle high degree of variability, that our approach can handle we articulate that our approach outperforms all the published popular methods.

[7] Complexity Reduced Face Detection Using Probability-Based Face Mask Profiteering and Pixel-Based Hierarchical-Feature Ad boosting:


The Ad boosting has attracted attention for its efficient face-detection performance. However, in the training process, the large number of possible Haar-like features in a standard sub-window becomes time consuming, which makes specific environment feature adaptation extremely difficult. This letter presents a two-stage hybrid face detection scheme using Probability-based Face Mask Pre-Filtering (PFMPF) and the Pixel-Based Hierarchical-Feature Ad boosting (PBHFA) method to effectively solve the above-mentioned problems in cascade Ad ad boosting. The two stages both provide far less training time than that of the cascade Ad boosting and thus reduce the computation complexity in face-detection tasks. In particular, the proposed PFMPF can effectively filter out more than 85% nonface in an image and the remaining few face candidates are then secondly filtered with a single PBHF Adaboost strong classifier. Given a M × N sub-window, the number of possible PBH features is simplified down to a level less than M × N, which significantly reduces the length of the training period by a factor of 1500. Moreover, when the two-stage hybrid face detection scheme are employed for practical face-detection tasks, the complexity is still lower than that of the integral-image based approach in the traditional Ad boosting method. Experimental results obtained using the gray ferret database show that the proposed two-stage hybrid face detection scheme is significantly more effective than Haar-like features.

[8] Face detection using the 3×3 block rank patterns of gradient magnitude images and a geometrical face model:


Face detection is required prior to various face-related applications. The objective of face detection is to determine whether or not there are any faces in an image and, if any, the location of each face is shown. Face detection in a natural scene image is challenging due to large variability of faceappearances. This paper proposes a face detection algorithm using the 3×3 block rank patterns of gradient magnitude images and a geometrical face model. The 3×3 block rank patterns are used to roughly classify whether the detected face candidate region contains a face or not. Finally, the face, if any, is detected by using a geometrical face model. Experimental results show that the proposed facedetection algorithm is insensitive to illumination in natural scene images.

[9] Face detection using one-class-based support vectors:


Almost all the proposed approaches regard face detection as a typical two-class pattern classification task, i.e., face pattern vs. non-face pattern, and learn face detector from face samples and non-facesamples. In practice, face pattern model can be established easily, while it is hard to gain a perfect non-face pattern model, for any pattern beyond face pattern (cat, plane, flower etc.) should belong to non-face pattern. In this paper, we propose a novel face detection approach based on one-class SVM (OCSVM), in which face detection is just considered to be a one-class pattern problem. Support vectors are used to model face pattern, and non-face patches in given images are rejected based on this model. In order to further improve performance, course-to-fine strategy is used in both the training anddetection procedure. Extensive experiments show that the proposed method has an encouraging performance.

[1௦] Face detection and geometric face normalization:


Face detection is a prerequisite step for face recognition and face analysis related applications. Aim of the face detection system is to identify and locate all faces regardless of their positions, scale, orientation, lighting conditions, expressions, etc. The faces which get detected under such conditions will affect the face recognition rate. Various approaches have been employed to detect faces in images. But as the face detection task is quite complex, each method is build in a precise context. The two main face detection approaches are: image based methods and geometrical based methods. We will focus on a detector which processes images very quickly while achieving high detection rates. This detection is based on a boosting algorithm called Ada Boost and the response of simple Haar-based features used by Viola and Jones. This paper presents a geometric normalization method based on eyes detection using Haar features and Ada Boost algorithm. The proposed method effectively normalizes the in-plane oriented faces and improves the face recognition rate.

[11] A face detection algorithm based on modified skin-color model:


Fast face detection is the important premise of biometrics, and it should be taken into account for its efficiency and resource cost in practical applications. Results can be quickly achieved for the traditional skin-color model of face detection but it can be affected in different lighting environments; we therefore improved the traditional skin-color model by experiments and apply the proposed model to design a fast eye location algorithm on frontal view face. Furthermore, we limited the distance of face and the camera and finally realized passive face detection related to the distribution of skin-color and the distance of two eyes. Different from the conventional methods, our algorithm makes full use of the relationship between the distance of two eyes and the distance between face and camera to assist in face detection, it devises a feasible way to promote efficiency in lip-reading and other non-specific face recognition applications. The face detection experiments under different lighting conditions also show that the proposed algorithm can achieve effective sensitivity and environmental robustness.

[12] Face detection using multi-modal features:


In this paper, propose efficient methods for reducing false detection rate in face detection procedure when images apply face recognition and face entertainment from web or personal photo. Previous detection methods focused on accurate face detection. So, false detection rate makes a problem when put to practical use. This problem can overcome to use additional information when face detection. Proper information is color. But real color-RGB information is not adequate for face detection. Because RGB information is sensitive to light variations. Thus need to transfer coordination properly to use. Transferred channels are defined with genetic algorithm. After that make a model of facial skin color to score face candidate is real face or not. Propose method can dramatic reduce false detection rate when face detection procedure.

[13] Study of Face Detection Algorithm for Real-time Face Detection System:


In this paper, the algorithm of face detection based on AdaBoost was studied, which is the multi-classifier cascade face detection algorithm and can attain real-time face detection in the system. It is a algorithm to automatically generate cascade classifier in the training process, which can effectively avoid the phenomenon of over trained. Compare to traditional AdaBoost face detection algorithm and feature subspace algorithm, the detection rate of the algorithm mentioned in this paper is not only higher nearly 10 percent, the detection velocity is also improved by more than one times. Also it is very robust to the illumination, pose etc variations and suitable for the real-time face detection system.

[14] Exploring the tradeoff between power and detection rate for a face detection algorithm:


Human face detection plays an important role in many applications such as human computer interface, video surveillance, face recognition, and face image database management. Recently, most researches mainly focus on developing a high detection rate algorithm to improve the performance. Since a high performance face detection algorithm requires complicated framework, it results high energy consumption and hard to implement in portable devices whose battery life’s are always restricted. Therefore, how to develop a face detection algorithm with both promising detection rate and low power dissipation has become a crucial research topic. In this paper, we present a comprehensive analysis of energy requirements in a wide range of the most used algorithms in the face detection program. The experimental results show the power dissipation and detection rate of seven different cases. Base on the results, we conclude and discuss various opportunities for realizing energy-efficient implementations of a face detection program.

[15] Initial experimental results of real-time variant pose face detection and tracking system:


A face detection system is a computer application for automatically detecting a human face from digital image or video frame. This paper presents a face detection system that used web camera to detect and track a face in real-time. To detect a face in the image, a simple method of skin color detection is used. By using color detection method in this project, the face can be segmented easily from the complex background. However, to detect a face in real-time is quite challenging especially when a faceis moving and the real-time environment has uneven illumination. This paper presents the preliminary result of face detection and tracking system, which is the system, detects a face that has different poses in a real-time situation, where the light condition is uneven. Here, to complete the detectionprocess, contour detection method is added so that the detection is more accurate. This system can be applied in many applications such as banking system to reduce the number of forgery, security system, and human-computer interaction (HCI).

[16] Temporal correlation and probabilistic prediction based face detection framework in real time environment:


Conventionally, the object detection algorithm proposed by Viola Jones (using Haar-like features, Integral image and AdaBoost algorithm) is implemented in majority of the face detection applications. In this article, an improvement to the application of Viola Jones algorithm in real time environment is presented by exploiting the analogy between video motion estimation and continuous object detection. Usingthe temporal correlation between successive frames of a real time input, various modifications improving the robustness and computational complexity of face detection are proposed. The proposed method focuses on reducing the search area for face detection on the basis of probabilistic prediction. In addition, approximation of minimum face size renders an improved performance. The modified FaceDetection Framework (FDF) is applied in two scenarios, canned video sequences from public databases and real time inputs from a low resolution camera, yielding improved results in both the cases.

[17] Face detection using combination of Neural Network and Adaboost:


High false positive face detection is a crucial problem which leads to low performance face recognition in surveillance system. The performance can be increased by reducing these false positives so that non-face can be discarded first prior to recognition. This paper presents a combination of two well known algorithms, Adaboost and Neural Network, to detect face in static images which is able to reduce the false-positives drastically. This method utilizes Haar-like features to extract the face rapidly using integral image. A cascade Adaboost classifier is used to increase the face detection speed. Due to using only this cascade Adaboost produces high false-positives, neural network is used as the final classifier to verify face or non-face. For a faster processing time, hierarchical Neural Network is used to increase the face detection rate. Experiments on four different face databases, which consist more than one thousand images, have been conducted. Results reveal that the proposed method achieves about 93.34% of detection rate and 0.34% of false-positives compared to original cascade Adaboost method which achieves about 98.13% of detection rate with 6.50% of false-positives. The processed images size is 240 × 320 pixels. Each frame is processed at about 2.25 sec which is slightslower than the original method, which only takes about 0.82 sec.

[18] Real-time face detection using Gentle AdaBoost algorithm and nesting cascade structure:


In this paper, a face detector based on Gentle AdaBoost algorithm and nesting cascade structure is proposed. Nesting cascade structure is introduced to avoid that too many weak classifiers in a cascade classifier will slow down the face detection speed of this cascade classifier. Gentle AdaBoost algorithm is used to train node classifiers on a Haar-like feature set to improve the generalization ability of the node classifier. Consequently, the face detection performance of the face detector is improved. Experimental results have proved that the proposed algorithm can significantly reduce the number of weak classifiers, increase the detection speed, and slightly raise the detection accuracy as well. On the CIF (352×288) video sequences, the average detection speed of the proposed face detector can achieve 125fps, which is superior to the state-of-the-art face detectors and completely satisfies the demand of real-time face detection.

[19] Combining Skin-Color Detector and Evidence Aggregated Random Field Models towards Validating Face Detection Results:


In this paper, a framework for validating any generic face detection algorithm's result is proposed. A two stage cascaded face validation filter is described that relies on a skin-color detector and on a facesilhouette structure modeler towards increasing face detection capacity of any face detection algorithm. While the skin-color detector combines a static skin-color and a dynamic background-color modeler, the face silhouette structure modeler incorporates an aggregate of random field models combined through a Dempster-Shafer framework of evidence merging. Together, the two modelers validate anyface subimage generated by face detection algorithms. Experiments conducted on FERET and on an in-house face database supports the claim for improved face detection results using the proposed filter. An extension of the same framework towards head pose estimation is also suggested.

[20] Real time head tracking and face and eyes detection:


This paper presents a novel algorithm to detect face and eyes in a reliable manner with a stereo camera. With the combination of stereo tracking method for object detection and tracking and structure-based method and example-based learning approach for face detection a real time and robust system is built. Human head detection is implemented with a stereo vision system, where a layered scenery is obtained from the depth image. With head-shoulder contour model applied to the disparity image, a fast head detection algorithm is developed. On the head detected, which also tells the rough scale of it, a hybrid face detection algorithm is proposed for face pattern detection and eyes location. By applying theface structure model and facial components model, face candidates in images are obtained for the later validating of face pattern. The support vector machine (SVM) is adopted here to select the good samples that form a classification boundary to classify face and non-face patterns.

[21] Face detection and tracking in color images using color centroids segmentation:


Human face detection plays an important role in many application areas such as video surveillance, human computer interface, face recognition, face search and face image database management etc. In human face detection applications, face region usually form an inconsequential part of images. Consequently, preliminary segmentation of images into regions that contain ldquonon-facerdquo objects and regions that may contain ldquofacerdquo candidates can greatly accelerate the process of human face detection. Color information based methods take a great attention, because colors have obviously character and robust visual cue for detection. This paper proposed a new method based on RGB color centroids segmentation (CCS) for face detection. This paper include two parts, first part is color image thresholding based on CCS and the second part is face detection based on region growing and facial features structure character combined method. The experimental results show the ideal thresholding result and better than the result of other color space analysis based thresholding methods. Proposed method can conquer the influence of different background conditions, position, scale instance and orientation in images from several photo collections and database; the effect is also better than existing skin color segmentation based methods.

[22] Rotational Invariant Face Detection on a Mobile Device:


In this paper we propose a rotation invariant face detection algorithm to detect the face on the mobile devices. Due to the performance limit on the mobile devices, complex face detection process must be avoided in order to generate a fast detection. Most proposed face detection algorithms are highly rotational variant, which means only upright face can be detected in the image, and those existed rotational invariant face detection algorithms are too complicated to be suitable for the mobile devices. Our algorithm can detect non-upright face in the image with an efficient process implemented on the mobile device (Google Nexus 7 tablet), and has different detection ranges adapting to different user choices. The experiment results reveal the validity of our proposed algorithm. Meanwhile, our algorithm can be easilyimplemented on other mobile systems, such as Windows Phone, iPhone and so on.

[23] A real time face tracking system using rank deficient face detection and motion estimation:


In this paper, we present a novel real-time face tracking system using rank deficient face detection. Real time computer vision systems need to be fast and efficient to be of practical application. Moreover, the system should be adaptable to diverse real life situations with varied availability of resources. We detect human face in an input grayscale image using a rank deficient face detection algorithm, which has been shown to be 4 to 6 times faster than unconstrained reduced set systems. The algorithm works on grayscale images and so easily adapts to change in ambient conditions like lighting conditions and different people, and is compatible with both color and grayscale cameras. One major problem in face tracking system used in environments with multiple candidate faces is which face to track and how to keep tracking a heuristically selected face without actually involving face authentication procedures. We have developed a highly flexible yet robust algorithm to deal with these problems. In case of multiple faces in the frame, the most prominent face is identified and automatically selected for tracking, and Webcam focused on the desired face. Motion estimation and compensation is then incorporated to ensure robust tracking, to minimize false detections, and for persistent tracking of the desired face. The system showed a robust 10 fps and a false detection rate of 0.6% on a 1.4 GHz Pentium IV workstation using an Intel CS110 Webcam. The principle used in this system can be applied to real time tracking of other objects, viz. cars, the human hand, etc. by using an appropriate training set. It can also be extended to provide face authentication and subsequent tracking of desired face.

[24] Face Detection Using Eigenface and Neural Network:


Face recognition is a popular research topic in the biometric identification area. Face detection and localization is the most important pre-processing module of a face recognition system. The purpose of face detection is to search and localize the positions of faces in an image in the varied background. In this paper, we propose a face detection system which combines the eigenface algorithm with the neural network. The candidate face region is examined to see if it is a face image block by analyzing the geometrical distribution of the edges in the region. In our experiments, the proposed face detectionsystem has a high face detection rate and less false detections.

[25] Design of face detection and tracking system:


Based on the face detection theory of Adaboost, incorporating with the program of Visual C++ in Windows, this paper designs a face detection system. This system realizes the function of face detection and increases detecting rate and speed by the combination of MIT face database and self-created face database, and the effective training of face database. Utilizing continuously adaptive mean shift (Camshift) algorithm to trace the face, this paper improves the Camshift algorithm. Utilizing the method of initialization template of Adaboost face detection, this paper combines the detection with tracing to increase the face tracing efficiency.

[26] Face Detection Using Radial Basis Function Neural Networks with Variance Spread Value:


This paper presents a face detection system using radial basis function neural networks with variance spread value. Face detection is the first step in face recognition system. The purpose is to localize and extract the face region from the background that will be fed into the face recognition system for identification. General preprocessing approach was used for normalizing the image and a radial basis function (RBF) neural network was used to distinguish between face and non-face images. RBF neural networks offer several advantages compared to other neural network architecture such as they can be trained using fast two stages training algorithm and the network possesses the property of best approximation. The output of the network can be optimized by setting suitable values of the center and spread of the RBF. In this paper, variance spread value will be used for every cluster where the value of spread will be calculated using algorithm. The performance of the RBFNN face detection system will be based on the detection rate, false acceptance rate (FAR) and the false rejection rate (FRR) criteria.

[27] A fast method of face detection in video images:


Face detection is the basis of all the face processing system, while in video the face detection problem has a more special significance. By studying the face detection based on Adaboost algorithm, this paper presents a fast and good robust face detection method. Firstly, the motion region which contains faces is obtained based on motion detection, excluding the background interference. Secondly, Adaboost algorithm is used to detect the face in the motion region and locate the face. The experiments show that this method can rapidly and accurately detect human faces.

[28] Face detection in complex background based on skin color features and improved AdaBoost algorithms:


Face detection is a very hot topic in research application of pattern recognition and computer vision. It is widely applied in artificial intelligence, video surveillance, identity authentication, human-machine interaction and so on. However, skin color detection has high false positive rate in complex background and AdaBoost algorithm was not satisfactory for detection of multi-pose and multi-face image. So a novel face detection method combined with skin color detection and an improved AdaBoost algorithm is proposed in this paper. First, it applies skin model segmentation and morphological operators to detect skin regions in the image. And according to the geometrical characteristics of the face, it screens the candidate face regions. Then by the improved classifiers in a cascade structure based on AdaBoost, it achieves more accurate promising regions of face. The experiment results show that this face detection algorithm improves the detection speed in both the quality of detection, and it can effectively reduce the error detection rate of single test method. This method has a good performance on image with complex background. Above all, this method has a certain theory value and practical value.

[29] A 0.64 mm Real-Time Cascade Face Detection Design Based on Reduced Two-Field Extraction:


Face detection is widely used in portable consumer handheld devices aimed at low area, low power, and high performance applications. The boosted cascade algorithm is one of the fastest face detection algorithms in use, but its hardware implementation requires a huge amount of SRAM to store the input data, integral image, and classifiers. This paper proposes a novel cascade face detection architecture based on a reduced two-field feature extraction scheme for faster integral image calculation and feature extraction. This scheme reduces the required memory for storing integral images by 75%, and employs multiple register files instead of a single SRAM to speed up the integral image updating and feature extraction processes. The reduced integral images have only 5% of the features of original images. Although this approach requires more weak classifiers, the proposed parallel cascade detection architecture reduces the average detection time for one feature to 63% that of the original. A 0.64 mm215 mw (@390 fps) boosted cascade face detection is implemented under the UMC 90-nm CMOS technology. Experimental results show that this face detection system can achieve a high face detection rate in processing 160 × 120 grayscale images at a speed of 390 fps.

[3௦] A Save-Fail mechanism for face detection validation:


A new automated mechanism named save-fail for validating face detection is proposed in this paper in order to determine whether or not there does exist faces in the candidate region which was located by the AdaBoost face detection algorithm. The mechanism is based on one simple fact, every face has two eyes and a nose, and human eyes and nose has specified features that can be used to distinguish them from the other objects; during the validation procedure, the mechanism first detects eyes and nose in the candidate region, then weights the eyes and nose detection results and normalizes these results to the designed range to accumulate the weights, at last, according to the relationship between the accumulated weight and the pre-defined threshold, determine whether or not there does exist faces. The experiments on CMU-MIT face set and practical testing set show that the mechanism is rather effective in the face detection system, to the positive samples, the mechanism greatly decreases the false detection and little correct detection is lost; to the negative samples, the mechanism both greatly decrease the false detection and improve the correction detection.

[31] Hybrid face detection system with robust face and non-face discriminability


While the face detection algorithm proposed by Viola and Jones has enough detection speed in still or video images and is applicable in practical applications, it is not reliable yet. That is, the number of false positives increases as background complexity increases. And sometimes it reports false positives, even for simple backgrounds. This problem can get the automatic face recognition systems (AFRSs) into trouble. In this paper, a hybrid system for face detection is introduced concerning itpsilas applicability in real world applications. Our approach is based on Viola and Jonespsilas work and uses Radial Basis Neural Network (RBFNN). The main characteristic of our approach is its robust face and non-face discriminability. Due to the extensive experiments, it is shown that the proposed system has decreased about 90% of false positives reported by the Viola and Jones algorithm.

[32] The system of face detection based on OpenCV


Face detection technology has widely attracted attention due to its enormous application value and market potential, such as face recognition and video surveillance system. Real-time face detection not only is one part of the automatic face recognition system but also is developing an independent research subject. So, there are many approaches to solve face detection. Here the modified AdaBoost algorithm based on OpenCV is presented, and experiments of real-time face detecting are also given through two methods of timer and dual-thread. The result shows that the method of face detection with dual-thread is simpler, smoother and more precise.

[33] Proposed FPGA Hardware Architecture for High Frame Rate (≫100 fps) Face Detection Using Feature Cascade Classifiers:


Face detection is the first and most crucial step in any face recognition systems with applications toface tracking, recognition and automatic surveillance. Gadabouts based face detection training methods have been proposed for producing a rapidly fast detection implementation compared to other methods using the integral image for evaluating a series of weak classifiers very fast. Current software version implementations can achieve about 15-25 frames per second (fps) with a tunable compromise between detection accuracy and speed. In this paper, a novel hardware architecture design on FPGA based on AdaBoost face training and detection algorithm for detecting faces in high resolution images at high frame rates (>100 fps) is proposed. The proposed architecture can evaluate each sub-window in a single clock cycle, which is the fastest possible speed land is limited by the H/W clock running speed. Thus in this proposed approach, detection speed is independent from the number of weak classifiers implemented. The proposed architecture is verified on Xilinx Virtex-II Pro FPGA platform where experimental results show that the speed and memory outperform previous approaches in literature, achieving 143 fps for image size of 640 by 480 pixels using a single scan window. Parallelizing the scan window can lead to double/triple this speed and is dependent on the gate capacity of the FPGA

[34] Efficient Face Detection with Segmentation and Feature-based Face Scoring in Surveillance Systems:


Facial images with low resolution in surveillance sequences are hard to detect with traditional approaches. An efficient face detection and face scoring technique in surveillance systems is proposed. It combines spirits of image-based face detection and essences of video object segmentation to filter out high-quality faces. The proposed face scoring technique, which is useful for surveillance video summary and indexing, includes four scoring functions based on feature extraction and is integrated by a neural network training system to select high-quality face. Experiments show that the proposed algorithm effectively extracts low-resolution human faces, which traditional face detection algorithm cannot handle well. It can also rank face candidates according to face scores, which determine face quality.

[35] Audio-visual face detection for tracking in a meeting room environment:


A key task in many applications such as tracking or face recognition is the detection and localisation of a subject's face in an image. This can still prove to be a challenging task particularly in low resolution or noisy images. Here we propose a robust method for face detection using both audio and visual information. We construct a dictionary learning based face detector using a set of distinctive and robust image features. We then train a support vector machine classifier using sparse image representations produced by this dictionary to classify face versus background. This is combined with the azimuth angle of the speaker produced by an audio localisation system to constrain the search space for the subject'sface. This increases the efficiency of the detection and localisation process by limiting the search area. However, more importantly, the audio information allows us to know a priori the number of subjects in the image. This greatly reduces the possibility of false positive face detections. We demonstrate the advantage of this proposed approach over traditional face detection methods on the challenging AV16.3 dataset.

[36] Face Detection and Tracking for Human Robot Interaction through Service Robot:


In our approach, a method is proposed to accomplish face detection and human tracking. In the realtime application such as robot system, the initialization and lost track problems often limit the application. The method which combines with face detection algorithm based on AdaBoost and the object tracking method using Kalman filter is proposed in this paper. To carry out the initialization problem, the global AdaBoost face detection (GAFD) algorithm is applied. It is powerful to detect multiple faces in the whole image. If there is a new face in image, a new face tracker and the new local AdaBoost face detection (LAFD) are both generated. The tracker will predict the possible area of theface appearing. According to the prediction of the tracker, partial image which is called region of interest (ROI) is obtained for LAFD. In the experimental result, our approach has been successfully implemented and test on service robot for human tracking. It is useful for the application of human-robot interaction

[37] Face Detection using DT-CWT on Spectral Histogram:


A method using DT-CWT spectral histograms and support vector machine (SVM) for face detection is proposed in this paper. In this face detection method, dual tree complex wavelets transform (DT-CWT) filter is used instead of Gabor filter. DT-CWT is a novel wavelet transform recently studied, which provides good directional selectivity in six different fixed orientations at different scales. It has limited redundancy for images and is much faster than Gabor transform to compute. Experimental results show that DT-CWT is a proper candidate to replaceGabor transform in face detection based on spectral histograms. By using an SVM trained on a set of 4000 faces aligned and 6000 non-face images produced by bootstrap algorithm, a robust classifying function for face and non-face pattern is obtained. The preliminary experimental results show that DT-CWT applied in face detection gives the satisfying performance. Several further improvements in computation time and performance are discussed

[38] Distribution-Based Face Detection using Calibrated Boosted Cascade Classifier:


This study presents a distribution-based face detection. The proposed face detection integrates information over image space and scale because the face can be recognized even if a small warp is applied. We must calibrate raw classifier outputs for information integration. Two salient contributions of this paper are calibration for a boosted cascade classifier and a distribution-based face detection. We first propose a new calibration method that converts the raw classifier outputs into posterior probabilities. This calibration method enables us to obtain a "Face Likelihood Distribution "from an input scene. The second contribution is the distribution-based face detection, which evaluates the local facelikelihood distribution over space and scale. In contrast, existing methods evaluate the face similarity value only at that point, which causes misdetection. Experimental results using 170input scenes, which might or might not include faces, show that the proposed method improves the detection rate by about 5%.

[39] Multi-pose Face Detection Using Facial Features and AdaBoost Algorithm:


Multi-pose face detection has been one of difficult and hot issues in face detection research nowadays, and urgently need to be solved in practical application. In this paper, a multi-pose face detectionalgorithm based on facial features and AdaBoost algorithm is introduced. Making full use of facial skin color information firstly, the most background regions can be quickly excluded and the skin color regions can be gotten. After detecting eyes and mouth, the face candidates are segmented according toface orientation decided by the geometric features of the eyes and mouth regions. At last, the facecandidates are verified by AdaBoost algorithm. The experimental results demonstrate that the algorithm can further improve the multi-pose face detection accuracy and is highly robust to facial expression and occlusion.

[40] Unsupervised learning for modulated view-based face detection system:


In order to realize an unsupervised learning for view-based face detection system, we introduce an architecture with separation of clustering features from detecting features. In this architecture, the number of detectors for individual "view" of faces is determined by the result of clustering autonomously. For autonomous clustering, the combination of Kohonen's self-organizing feature map (SOM) and a novel cluster determination algorithm is introduced. This cluster determination algorithm allows us to find gaps between clusters by comparing the distance from a particular data to adjacent clusters. Therefore, it can define the similarity of face "view" to represent the view-based features in case of face detection. For feature detection, a non-linear subspace model based on kernel method is used. Our architecture self-organizes feature detectors corresponding to face "view"; the face detectionsystem shows good face detection performance under a wide variety of lighting conditions.

[41] Face Detection Based on Facial Features and Linear Support Vector Machines:


Face detection is a complicated and significant problem in pattern recognition and has wide application. This paper proposes a fast face detection algorithm based on facial features and linear Support Vector Machines (LSVM). First, using of skin color information, the algorithm quickly excludes most background regions from the images primarily leaving the skin color regions. Then we use LSVM to separate more non-face regions from the remaining regions, for exiting big differences between the faceregions and non-face regions. Finally, we identify the face candidates by detecting eyes and mouth. The experimental results demonstrate that the algorithm can further improve the detection accuracy and lower false detection rate and greatly speed up the detection rate.

[42] Face Detection Based on Skin Colour Model and Geometry Features:


Face detection has been a key issue for computer vision. It is essential to many applications, for example, human face synthesis, face tracking, pose estimation, facial expression recognition and object oriented image coding. Detection and localization of human faces is still a challenge because of several difficulties, such as variable face orientations, different face sizes, partial occlusions of faces, and changeable lighting conditions. In this paper, we present an algorithm to face detection in video sequences. It is a hierarchical approach, which integrates a skin color model and gradient features. Results show that the method improved the accuracy of front face detection.

[43] Face detection using distribution-based distance and support vector machine:


This paper presents a novel face detection method by applying distribution-based distance (DBD) measure and support vector machine (SVM). The novelty of our DBD-SVM method comes from the integration of discriminating feature analysis, face class modeling, and support vector machine for face detection. First, the discriminating feature vector is defined by combining the input image, its 1-D Haar wavelet representation, and its amplitude projections. Then the DBD-SVM method statistically models the face class by applying the discriminating feature vectors and defines the distribution-based distance measure. Finally, based on DBD and SVM, three classification rules are applied to separate faces and no faces. Experiments using images from the MIT-CMU test sets show the feasibility of our new face detection method. In particular, when using 92 images (containing 282 faces) from the MIT-CMU test sets, our DBD-SVM method achieves 98.2% correct face detection accuracy with 2 false detections, a performance comparable to the state-of-the-art face detection methods, such as the Schneiderman-Kanade's method.

[44]Study on Face Detection Algorithm Based on Skin Color Segmentation and AdaBoost Algorithm:


Because skin color detection had high false detecting rate in color image with complex background and the AdaBoost face detection algorithm does not work well in multi-pose and multi-face images. In this paper, a new face detection algorithm is introduced combinedthe skin color segmentation and the AdaBoost algorithm. The simulated experiment results show that this algorithm mentioned in this paper, not only reduces the false detecting rate and missing rate, but also obviously improves the detectionspeed. And the algorithm can be applied to more complex background face detection which has multi-pose and multi-face of large images.

[45] Face detection under variable lighting based on resample by face relighting:


Different environment illuminations have a great impact on face detection. We present a solution based on face relighting technology. The basic idea is that there exists nine harmonic images that can be derived from a 3D model of a face, and by which we can estimate the illumination coefficient of any facesample. Using an illumination radio image, we can produce images under new lighting conditions. To detect the faces under certain lighting condition, we relight the original face samples to get more new faces under different kinds of possible lighting condition, and add them to the training set. Our experimental results on support vector machine (SVM) turns out that the relighting subspace is effective in face detection under various lighting conditions. Moreover, if we relight original face samples to the new samples under different illuminations, the collected example sets are multiplied. We use the expanded database to train an AdaBoost-based face detector and test it on the MIT+CMU frontal facetest set. The experimental results show that the data collection can be efficiently speeded up by the proposed methods. The later experiment also verifies the generalization capability of the proposed method.

[46] A Bayesian discriminating features method for face detection:


This paper presents a novel Bayesian discriminating features (BDF) method for multiple frontal face detection . The BDF method, which is trained on images from only one database, yet works on test images from diverse sources, displays robust generalization performance. The novelty of this paper comes from the integration of the discriminating feature analysis of the input image, the statistical modeling of face and no face classes, and the Bayes classifier for multiple frontal face detection. First, feature analysis derives a discriminating feature vector by combining the input image, its 1D Harr wavelet representation, and its amplitude projections. While the Harry wavelets produce an effective representation for object detection, the amplitude projections capture the vertical symmetric distributions and the horizontal characteristics of human face images. Second, statistical modeling estimates the conditional probability density functions, or PDFs, of the face and no face classes, respectively. While the face class is usually modeled as a multivariate normal distribution, the no face class is much more difficult to model due to the fact that it includes "the rest of the world." The estimation of such a broad category is, in practice, intractable. However, one can still derive a subset of the no faces that lie closest to the face class, and then model this particular subset as a multivariate normal distribution.

[47] Face Detection Based on Fuzzy Granulation and Skin Color Segmentation:


Face Detection is the process of determining the face location, size and number. Robust face detection in complex background is a challenging task. In this paper, a novel method based on skin color segmentation and classification with fuzzy information granulation (FIG) for robust and fast face detection in color images is proposed. In this method, firstly we use skin color segmentation to extract face candidates. After normalization of candidate faces, a fuzzy information granulation based classifier is used to classify face or non-face patterns. Experimental results show that this method is effective and fast with high accuracy in detecting faces in color images.

[48] A direct approach for face detection on omnidirectional images:


Cat dioptric sensors offer abilities unexploited so far. This is especially true for face detection, and more generally, object detection. This paper presents our results of a direct approach to tackle face detection on cat dioptric images. Despite no geometrical transformations, we are able to successfully apply our detector on distorted images. We expose a new method to synthesize large omnidirectional images database. Inspired from regular face detection training schemes, our method makes use of newly introduced polygonal Haar-like features. First tests demonstrated that our approach gives good performance and at the same time speeds up the detection process.

[49] Rotation Invariant Multi-View Colour Face Detection Based on Skin Colour and Ad boost Algorithm:


As the training of Adaboost is complicated or does not work well in the multi-view face detection with large plane-rotated angle, this paper proposes a rotation invariant multi-view color face detection method combining skin color segmentation and multi-view Adaboost algorithm. First the possible faceregions is fast detected by skin color table, and the skin-color background adhesion of face is separated by color clustering. After region merging, the candidate face region is received. Then the face direction in plane is calculated by K-L transform, and finally the candidate face region corrected by rotating is scanned by multi-view Adaboost classifier to locate the face accurately. The experiments show that the method can effectively detect the plane large-angle multi-view face image which the conventional Adaboost can not do. It can be effectively applied to the cases of multi-view and multi-face image with complex background.

[50] An Improved Face Detection Method in Low-resolution Video:


In this study, an efficient face detection method is proposed for low-resolution video. The cascaded face detector proposed by Viola can achieve real-time detection and a high detection rate. However, the motion blur of the face images in the low-resolution video usually exists. The detection rate in low-resolution video is lower than that in static images because the training set in the Adaboost algorithm only considers about normal face images. Therefore, the enhanced training set which contains the normal face images and the motion blurred face images is used to improve the detection rate. The simulation results show that the face images in low-resolution video can be efficiently extracted.

More About Face Detection:

Face recognition is used for two primary tasks:

Verification (one-to-one matching): When presented with a face image of an unknown individual along with a claim of identity, ascertaining whether the individual is who he/she claims to be.

Identification (one-to-many matching):

Given an image of an unknown individual, determining that person’s identity by comparing (possibly after encoding) that image with a database of (possibly encoded) images of known individuals.


Holistic approaches attempt to identify faces using global representations, i.e., descriptions based on the entire image rather than on local features of the face. These schemes can be subdivided into two groups: statistical and AI approaches. An overview of some of the methods in these categories follows.

Core Scenarios:

• In a video-conference scenario, a passive camera with a wide angle lens provides a global view of the persons seated around a conference table. An active camera provides a close-up of whichever person is speaking. Face detection, operating on the wide-angle images, can be used to estimate the location of conference participants around the table. When a participant speaks, the high resolution camera can zoom onto the face of the speaker and hold the face in the center of the image. • In the real-world usage of video surveillance, multiple cameras are utilized in a wide spectrum of applications for the purpose of monitoring. As a result, multiple monitors are used to display the various output video streams. Manual monitoring of multiple screens is tedious and operators are prone to fatigue. Face detection can help to reduce the data from multiple screens into one main screen for monitoring

DCT Based Face Recognition paper has been submitted on this important topic:

The next generation of algorithms were based on the overall appearance of the face. They are usually based on projecting the face image into a face space and using classification algorithms in this face space for computing similarities between faces of these techniques is based on computing the DCT spectrum of HMMs are build the face image and then using the entire spectrum of just a part of it for classification. The classification algorithms can vary from simple distances between DCT coefficients vectors or. more complex algorithms employing neural network techniques [7], or support vector machines [5,8]. We remark that as most digital cameras feature hardware to generate a local DCT. transform on image data that such techniques can be particularly advantageous with respect to computational efficiency for in camera implementation

Hidden Markov Model Techniques

One possible alternative to embedded face recognition that we tested is a 2D variant of Hidden Markov Models (HMMs) . One called embedded HMMs (EHlMM). The main advantage of the is that the models for each person independentlys. So every time we want to add a new person to the collection we just have to add a new model without modifying the other models The observations needed to train the models are the DCT coefficients which are fast to compute, and the wavelet coefficients, which are more robust to small variations By proper training these models can cope with face variations which is very important when working with consumer.

KaaShiv InfoTech offers world class Final Year Project for BE, ME, MCA ,MTech, Software engineering and other students in Anna Nagar, Chennai.

internship in chennai

Website Details:

Inplant Training:




Final Year Projects: