Статьи журнала - International Journal of Image, Graphics and Signal Processing
Все статьи: 1110
Face Recognition Based on Principal Component Analysis
Статья научная
The purpose of the proposed research work is to develop a computer system that can recognize a person by comparing the characteristics of face to those of known individuals. The main focus is on frontal two dimensional images that are taken in a controlled environment i.e. the illumination and the background will be constant. All the other methods of person's identification and verification like iris scan or finger print scan require high quality and costly equipment's but in face recognition we only require a normal camera giving us a 2-D frontal image of the person that will be used for the process of the person's recognition. Principal Component Analysis technique has been used in the proposed system of face recognition. The purpose is to compare the results of the technique under the different conditions and to find the most efficient approach for developing a facial recognition system
Бесплатно
Face Recognition Based on Texture Features using Local Ternary Patterns
Статья научная
Face recognition is one of the important and popular visual recognition problem due to its challenging nature and its diverse set of applications. That's why face recognition is attracted by many researchers. Methods based on Local Binary Pattern (LBP) are widely used for face recognition in the literature, and it is sensitive to noise. To address this present paper utilized the powerful local texture descriptor that is less sensitive to noise and more discriminant in uniform regions called as Local Ternary Pattern (LTP). The Uniform Local Binary Pattern (ULBP) derived on LBP treats a large set of LBP under one label called as miscellaneous. This may result some loss of information on LBP and LTP based methods. To address this two Prominent LBP (PLBP) are derived, namely PLBP-Low (L) and PLBP-High (H) on LTP. Based on this the present paper derived eight texture features on facial images. A distance function is used on proposed texture features for effective face recognition. To eliminate most of the effects of illumination changes that are present in human face an efficient preprocessing method is used that preserves the significant appearance details that are needed for face recognition. The present method is experimented on Yale, Indian and American Telephone and Telegraph Company (AT&T) Olivetti Research Laboratory (ORL) data bases and it has given state-of-the-art performance on the three popular datasets.
Бесплатно
Статья научная
In this paper, we introduce a face recognition algorithm based on doubly truncated multivariate Gaussian mixture model with DCT under logarithm domain. In face recognition, the face image is subject to the variation of illumination. The effect of illumination cannot be avoided by mere consideration of DCT coefficients as feature vector. The illumination effect can be minimized by utilizing DCT coefficients under logarithm domain and discarding sum of the DCT coefficients which represents the illumination in the face image. Here, it is assumed that the DCT coefficients under logarithm domain after adjusting the illumination follow a doubly truncated multivariate Gaussian mixture model. The truncation on the feature vector has a significant influence in improving the recognition rate of the system using EM algorithm with K-means or hierarchical clustering, the model parameters are estimated. A face recognition system is developed under Bayesian frame using maximum likelihood. The performance of the system is demonstrated by using the databases namely, JNTUK and Yale and comparing it’s performance with the face recognition system based on GMM. It is observed that the proposed face recognition system outperforms the existing systems.
Бесплатно
Face Recognition System based on Convolution Neural Networks
Статья научная
Face Recognition plays a major role in the new modern information technology era for security purposes in biometric modalities and has still various challenges in many applications of computer vision systems. Consequently, it is a hot topic research area for both industrial and academic environments and was developed with many innovative ideas to improve accuracy and robustness. Therefore, this paper proposes a recognition system for facial images by using Deep learning strategies to detect a face, extract features, and recognize. The standard facial dataset, FEI is used to prove the effectiveness of the proposed system and compare it with the other previous research works, and the experiments are carried out for different detection methods. The results show that the improved accuracy and reduce time complexity can provide from this system, which is the advantage of the Convolution Neural Network (CNN) than other some of the previous works.
Бесплатно
Face Recognition Using Modified Histogram of Oriented Gradients and Convolutional Neural Networks
Статья научная
We are aiming in this work to develop an improved face recognition system for person-dependent and person-independent variants. To extract relevant facial features, we are using the convolutional neural network. These features allow comparing faces of different subjects in an optimized manner. The system training module firstly recognizes different subjects of dataset, in another approach, the module processes a different set of new images. Use of CNN alone for face recognition has achieved promising recognition rate, however many other works have showed declined in recognition rate for many complex datasets. Further, use of CNN alone exhibits reduced recognition rate for large scale databases. To overcome the above problem, we are proposing a modified spatial texture pattern extraction technique namely modified Histogram oriented gradient (m-HOG) for extracting facial image features along three gradient directions along with CNN algorithm to classify the face image based on the features. In the preprocessing stage, the face region is captured by removing the background from the input face images and is resized to 100×100. The m-HOG features are retrieved using histogram channels evenly distributed between 0 and 180 degrees. The obtained features are resized as a matrix having dimension 66×198 and which are passed to the CNN to extract robust and discriminative features and are classified using softmax classification layer. The recognition rates obtained for L-Spacek, NIR, JAFFE and YALE database are 99.80%, 91.43%, 95.00% and 93.33% respectively and are found to be better when compared to the existing methods.
Бесплатно
Face Super Resolution: A Survey
Статья научная
Accurate recognition and tracking of human faces are indispensable in applications like Face Recognition, Forensics, etc. The need for enhancing the low resolution faces for such applications has gathered more attention in the past few years. To recognize the faces from the surveillance video footage, the images need to be in a significantly recognizable size. Image Super-Resolution (SR) algorithms aid in enlarging or super-resolving the captured low-resolution image into a high-resolution frame. It thereby improves the visual quality of the image for recognition. This paper discusses some of the recent methodologies in face super-resolution (FSR) along with an analysis of its performance on some benchmark databases. Learning based methods are by far the immensely used technique. Sparse representation techniques, Neighborhood-Embedding techniques, and Bayesian learning techniques are all different approaches to learning based methods. The review here demonstrates that, in general, learning based techniques provides better accuracy/ performance even though the computational requirements are high. It is observed that Neighbor Embedding provides better performances among the learning based techniques. The focus of future research on learning based techniques, such as Neighbor Embedding with Sparse representation techniques, may lead to approaches with reduced complexity and better performance.
Бесплатно
Face recognition based on cross diagonal complete motif matrix
Статья научная
To extract local features efficiently Jhanwar et al. proposed Motif Co-occurrence Matrix (MCM) [23] in the literature. The Motifs or Peano Scan Motifs (PSM) is derived only on a 2*2 grid. The PSM are derived by fixing the initial position and this has resulted only six PSM’s on the 2*2 grid. This paper extended this ap-proach by deriving Motifs on a 3*3 neighborhood. This paper divided the 3*3 neighborhood into cross and diag-onal neighborhoods of 2*2 pixels. And on this cross and diagonal neighborhood complete Motifs are derived. The complete Motifs are different from initial Motifs, where the initial PSM positions are not fixed. This complete Motifs results 24 different Motifs on a 2*2 gird. This paper derived cross diagonal complete Motifs matrix (CD-CMM) that has relative frequencies of cross and diagonal complete Motifs. The GLCM features are de-rived on cross diagonal complete Motifs texture matrix for efficient face recognition. The proposed CD-CMM is evaluated face recognition rate on four popular face recognition databases and the face recognition rate is compared with other popular local feature based methods. The experimental results indicate the efficacy of the proposed method over the other existing methods.
Бесплатно
Face recognition based texture analysis methods
Статья научная
A unimodal biometric system based Local Binary Pattern (LBP) and Gray Level Co-occurrence Matrix (GLCM) is developed to recognize the facial of 40 subjects. The matching process is implemented using three classifiers: Euclidean distance, Manhattan distance, and Cosine distance. The maximum accuracy (100%) is satisfied when GLCM and LBP are applied with Euclidean distance. The accuracy result of these two methods is advanced the Principle Component Analysis (PCA) and Fourier Descriptors (FDs) recognition rate. The ORL database is considered for constructing the proposed biometric system.
Бесплатно
Facial Expression Recognition Based on Features Derived From the Distinct LBP and GLCM
Статья научная
Automatic recognition of facial expressions can be an important component of natural human-machine interfaces; it may also be used in behavioural science and in clinical practice. Although humans recognise facial expressions virtually without effort or delay, reliable expression recognition by machine is still a challenge. This paper, presents recognition of facial expression by integrating the features derived from Grey Level Co-occurrence Matrix (GLCM) with a new structural approach derived from distinct LBP’s (DLBP) ona 3 x 3 First order Compressed Image (FCI). The proposed method precisely recognizes the 7 categories of expressions i.e.: neutral, happiness, sadness, surprise, anger, disgust and fear. The proposed method contains three phases. In the first phase each 5 x 5 sub image is compressed into a 3 x 3 sub image. The second phase derives two distinct LBP’s (DLBP) using the Triangular patterns between the upper and lower parts of the 3 x 3 sub image. In the third phase GLCM is constructed based on the DLBP’s and feature parameters are evaluated for precise facial expression recognition. The derived DLBP is effective because it integrated with GLCM and provides better classification performance. The proposed method overcomes the disadvantages of statistical and formal LBP methods in estimating the facial expressions. The experimental results demonstrate the effectiveness of the proposed method on facial expression recognition.
Бесплатно
Facial Image Super Resolution Using Weighted Patch Pairs
Статья научная
A challenging field in image processing and computer graphics is to have higher frequency details by super resolving facial images. Unlike similar papers in this field, this paper introduces a practical face hallucinating approach with higher quality output images. The image reconstruction was based on a set of high and low resolution image pairs. Each image is divided into defined patches with overlapped regions. A patch from a defined location is removed from the low resolution (LR) input image and is compared with the LR patches of the training images with the same location. Each defined LR patch has a defined high resolution (HR) patch. Based on the Euclidean distance comparison, each patch of every single image in the training images database receives a specific weight. This weight is transferred to its relevant HR patch identically. The sum of the gained weights for one specific location of a patch is equal to unity. The HR output image is constructed by integrating the HR hallucinated patches.
Бесплатно
Facial expression recognition by holistic and geometrically integrated subspace
Статья научная
This paper demonstrates mainly on feature extraction by analytic and holistic methods and proposes a novel approach for feature level fusion for efficient expression recognition. Gabor filter magnitude feature vector is fused with upper part geometrical features and phase feature vector is fused with lower part geometrical features respectively. Both these high dimensional feature dataset has been projected into low dimensional subspace for de-correlating the feature data redundancy by preserving local and global discriminative features of various expression classes of JAFFE, YALE and FD databases. The effectiveness of subspace of fused dataset has been measured with different dimensional parameters of Gabor filter. The experimental results reveal that performance of the subspace approaches for high dimensional proposed feature level fused dataset compared with state of art approaches.
Бесплатно
Facial expressions recognition in thermal images based on deep learning techniques
Статья научная
Facial expressions are undoubtedly the best way to express human attitude which is crucial in social communications. This paper gives attention for exploring the human sentimental state in thermal images through Facial Expression Recognition (FER) by utilizing Convolutional Neural Network (CNN). Most traditional approaches largely depend on feature extraction and classification methods with a big pre-processing level but CNN as a type of deep learning methods, can automatically learn and distinguish influential features from the raw data of images through its own multiple layers. Obtained experimental results over the IRIS database show that the use of CNN architecture has a 96.7% recognition rate which is high compared with Neural Networks (NN), Autoencoder (AE) and other traditional recognition methods as Local Standard Deviation (LSD), Principle Component Analysis (PCA) and K-Nearest Neighbor (KNN).
Бесплатно
Fast 3D Volume Super Resolution Using an Analytical Solution for l2-l2 Problems
Статья научная
In Endodontics, dentists need a good visualization of dental root canals as found in Cone Beam Computed Tomography (CBCT) dental volumes to diagnose and prevent the development of some anomalies. These CBCT dental volumes, however, suffer from low resolution. In order, to enhance their resolution, we need to apply a super-resolution technique. In this paper, we propose a new 3D super resolution algorithm based on a linear model, consisting of a blurring operator and a decimation operator, which is an extension of Zhao’s work [1] in 3D, taking the low-resolution volume as an input and producing the high-resolution volume as an output. We present a generalization of the 2D Super-Resolution problem into a 3D Super- Resolution problem as we apply it to 3D dental volume. Our new Super-Resolution algorithm as applied to dental CBCT volumes is a direct method aiming to get the exact solution with a short computation time. Results show an improvement in the resolution of the CBCT in a short time in comparison with Zhao’s work, which was applied to CBCT dental volumes slice by slice, [2]
Бесплатно
Fast Encryption Scheme for Secure Transmission of e-Healthcare Images
Статья научная
E-healthcare systems (EHSD), medical communications, digital imaging (DICOM) things have gained popularity over the past decade as they have become the top contenders for interoperability and adoption as a global standard for transmitting and communicating medical data. Security is a growing issue as EHSD and DICOM have grown more usable on any-to-any devices. The goal of this research is to create a privacy-preserving encryption technique for EHSD rapid communication with minimal storage. A new 2D logistic-sine chaotic map (2DLSCM) is used to design the proposed encryption method, which has been developed specifically for peer-to-peer communications via unique keys. Through the 3D Lorenz map which feeds the initial values to it, the 2DLSCM is able to provide a unique keyspace of 2544 bits (2^544bits) in each go of peer-to-peer paired transmission. Permutation-diffusion design is used in the encryption process, and 2DLSCM with 3DLorenz system are used to generate unique initial values for the keys. Without interfering with real-time medical transmission, the approach can quickly encrypt any EHSD image and DICOM objects. To assess the method, five distinct EHSD images of different kinds, sizes, and quality are selected. The findings indicate strong protection, speed, and scalability when compared to existing similar methods in literature.
Бесплатно
Статья научная
In the present day real time applications of visual object tracking in surveillance, it has become extremely complex, time consuming and tricky to do the tracking when there are occlusions are present for small duration or for longer time and also when it is done in outdoor environments. In these conditions, the target to be tracked can be lost for few seconds and that should be tracked as soon as possible. As from the literature it is observed that particle filter can be able to track the target robustly in different kinds of background conditions, and it’s robust to partial occlusion. However, this tracking cannot recover from large proportion of occlusion and complete occlusion, to avoid this condition, we proposed two new algorithms (modified kalman and modified particle filter) for fast tracking of objects in the presence of occlusions. We considered the complete occlusion of tracking object and the main objective is how fast the system is able to track the object after the occlusion is crossed. From the experimental results, it is observed that the proposed algorithms have shown good improvement in results compared to the traditional methods.
Бесплатно
Fast Width Detection in Corridor Using Hough Transform
Статья научная
For many robotics and smart car applications it is vitally important to calculate the width. The present paper proposes a new approach for finding the width of a corridor within a constructed image frame that would keep a robot on a safe track away from walls. The main advantage of this approach is less computation time and hence faster response for path recognition. In this new approach, the Hugh Transform technique is also used as the basis of the provided algorithm. Within the determination of corridor width, in order to avoid the accident in the future researches, some approaches such as identify open space, modeling and reconstruction of three-dimensional space, can also be used.
Бесплатно
Fast and Accurate Classification F and NF EEG by Using SODP and EWT
Статья научная
Removing the brain part, as the epilepsy source attack, is a surgery solution for those patients who have drug resistant epilepsy. So, the epilepsy localization area is an essential step before brain surgery. The Electroencephalogram (EEG) signals of these areas are different and called as focal (F) whereas the EEG signals of other normal areas are known as non-focal (NF). Visual inspection of multi-channels for F EEG detection is time-consuming along with human error. In this paper, an automatic and adaptive method is proposed based on second order difference plot (SODP) of EEG rhythms in empirical wavelet transform (EWT) domain as an adaptive signal decomposition. SODP provides the data variability rate or gives a 2D projection for rhythms. The feature vector is obtained using the central tendency measure (CTM). Finally, significant features, chosen by Kruskal–Wallis statistical test, are fed to K nearest neighbor (KNN) and support vector machine (SVM) classifiers. The achieved results of the proposed method in terms of three objective criteria are compared with state-of-the-art papers demonstrating an outstanding algorithm here in.
Бесплатно
Fast generation of image’s histogram using approximation technique for image processing algorithms
Статья научная
The process of generating histogram from a given image is a common practice in the image processing domain. Statistical information that is generated using histograms enables various algorithms to perform a lot of pre-processing task within the field of image processing and computer vision. The statistical subtasks of most algorithms are normally effectively computed when the histogram of the image is known. Information such as mean, median, mode, variance, standard deviation, etc. can easily be computed when the histogram of a given dataset is provided. Image brightness, entropy, contrast enhancement, threshold value estimation and image compression models or algorithms employ histogram to get the work done successfully. The challenge with the generation of the histogram is that, as the size of the image increases, the time expected to traverse all elements in the image also increases. This results in high computational time complexity for algorithms that employs the generation histogram as subtask. Generally the time complexity of histogram algorithms can be estimated as O(N2) where the height of the image and its width are almost the same. This paper proposes an approximated method for the generation of the histogram that can reduce significantly the time expected to complete a histogram generation and also produce histograms that are acceptable for further processing. The method can theoretically reduce the computational time to a fraction of the time expected by the actual method and still generate outputs of acceptable level for algorithms such as Histogram Equalization (HE) for contrast enhancement and Otsu automatic threshold estimation.
Бесплатно
Feature Based Image Mosaic Using Steerable Filters and Harris Corner Detector
Статья научная
Image mosaic is to be combine several views of a scene in to single wide angle view. This paper proposes the feature based image mosaic approach. The mosaic image system includes feature point detection, feature point descriptor extraction and matching. A RANSAC algorithm is applied to eliminate number of mismatches and obtain transformation matrix between the images. The input image is transformed with the correct mapping model for image stitching and same is estimated. In this paper, feature points are detected using steerable filters and Harris, and compared with traditional Harris, KLT, and FAST corner detectors.
Бесплатно
Feature Selection based Breast Cancer Prediction
Статья научная
Breast cancer is one of the main causes of mortality for women around the world. Such mortality rate could be reduced if it is possible to diagnose breast cancer at the primary stage. It is hard to determine the causes of this disease that may lead to the development of breast cancer. But it is still important in predicting the probability of cancer. We can assess the likelihood of occurrence of breast cancer using machine learning algorithms and routine diagnosis data. Although a variety of patient information attributes are stored in cancer datasets not all of the attributes are important in predicting cancer. In such situations, feature selection approaches can be applied to keep the pertinent feature set. In this research, a comprehensive analysis of Machine Learning (ML) classification algorithms with and without feature selection on Wisconsin Breast Cancer Original (WBCO), Wisconsin Diagnosis Breast Cancer (WDBC), and Wisconsin Prognosis Breast Cancer (WPBC) datasets is performed for breast cancer prediction. We employed wrapper-based feature selection and three different classifiers Logistic Regression (LR), Linear Support Vector Machine (LSVM), and Quadratic Support Vector Machine (QSVM) for breast cancer prediction. Based on experimental results, it is shown that the LR classifier with feature selection performs significantly better with an accuracy of 97.1% and 83.5% on WBCO and WPBC datasets respectively. On WDBC datasets, the result reveals that the QSVM classifier without feature selection achieved an accuracy of 97.9% and these results outperform the existing methods.
Бесплатно
