Transform based face recognition with partial and full feature vector using DCT and Walsh transform

Objective: To identify optimum features from the initial feature set to classify ship ISAR images with good accuracy. The optimum feature selection brings down the computational complexity and also removes the redundant features in the feature vector. Methodology: Colour moments computed from R, G, B, RG, GB, RB component images are taken as one feature vector. The average wave energy values of horizontal, vertical and diagonal details of images decomposed up to five levels using wavelet transform are considered as second feature vector. The ship ISAR images measured with high resolution Radar are used as data set to test the classification performance of the two feature sets, which is explored uniquely in this work. Genetic algorithm and particle swarm optimization techniques are used to find the optimum feature sets. Probabilistic neural network is used to compare the classification accuracy achieved with full length feature vectors and optimum feature vectors of the two considered feature sets. Findings: Data set of two ships is taken. The data of each ship is divided in to two parts training data and test data. The feature vectors computed from training set are used to train the classifier and that computed from test set are used to test the classification accuracy. The number of variables (number of features in the feature vector) of optimization algorithms Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) is varied to find the optimum feature set to classify the ship ISAR image data. The function that computes the classification accuracy is used as the objective function of the optimization algorithms. From the initial feature vector of thirty colour moments a set of four features is identified that gives 100% accuracy for both the targets considered. From starting set of fifteen average energy levels, a set of four energy levels identified by PSO and the set of three energy levels are found to be more effective in classification of the ship ISAR images with classification accuracy of 66.6% for the first target and 75% for the second target. Both the optimization techniques performed in similar way. Other optimization algorithms can be explored to find feature combinations that classify ship ISAR images with better accuracy. Application: Classification based on ISAR images has application in automatic non-cooperative target recognition, development of monitoring systems for surveillance along transportation corridors etc. to avoid illegal immigration. Future Work: Different classifiers Self Organizing Maps, Support Vector Machines etc. can be used to study the possibility of obtaining optimum solution in smaller time. Keywords: Classification, Feature Selection, ISAR image, Optimum Feature Set, Optimization Techniques

This paper describes in detail about 2D and 3D face and video face databases developed in the year from 1992 to 2009. The still face database discussed here includes: Our Laboratory Database of Faces (ORL), Yale face database, Yale face databaseB(B stands for Big), Yale face database extendedB, FERET(Face Recognition Technology) face database and CASPEAL is a large scale Chinese face database with varying PEAL(Pose Expression Accessory, and Lighting). The 3D face databases like MIT-CBCL (Centre for Biological and Computational Learning) face recognition database, Georgia Tech face database, Cohn-Kanade Action Unit Face database and 3D BFM (Basel Face Model) are also discussed in this paper. In CK AU, CK+AU(Cohn-Kanade Action Unit) coded facial expression database, a 3D model is developed from a 2D model which is obtained from the captured video of different subjects. The other Face databases and development of new Face databases are discussed at the end of this paper. The application of proposed concept is to develop a new small size face database which includes all the important criteria included in the previous face and video face databases.

In Iris recognition, the identification and authentication of an individual is carried out by analysis of unique patterns of iris. This paper presents, extracting the unique features from the iris images. The feature extraction is done by using concept of energy compaction. Palacky University Database is used as a test bed for proposed iris recognition technique, which contains 384 iris images of 64 persons, with 3 images for left eye and 3 images for right eye of each person. For the proposed iris recognition technique, self mutated hybrid wavelet transform of Cosine-Walsh, Cosine-Kekre and Cosine-Sine are used to generate transformed iris images. Considering all the coefficients for 100% of energy, feature vector of transformed iris image is generated, thus the size of feature vector becomes extremely large. For proposed Iris recognition technique, 99%, 98%, 97% and 96% of partial energies are considered, thus the number of coefficients considered to generate the feature vector are very less and the size of feature vector reduces severely. Accuracy using partial energies is high as compared to 100% of energy. The self mutated hybrid wavelet transform gives improvement in Genuine Acceptance Rate (GAR) and faster recognition. The best result is obtained for Cosine-Walsh self mutated hybrid wavelet transform as compared to other combinations.

Digital images can be easily manipulated due to availability of powerful image processing software. Passive authentication as a common challenge method of digital image authentication is extensively used to detect the copy-move forgery images. In this paper, a passive authentication scheme is proposed to authenticate copy-move forgery based on discrete cosine transform (DCT). At the feature extraction step, DCT is applied to image blocks and makes use of the means of DCT coefficients to represent image blocks. The size of feature vectors are optimized. At the matching step, a set number of packages is used to store the feature vectors. The similar blocks can be found by comparing the feature vectors that are contained in adjacent packages. The experimental results demonstrate that the proposed scheme can locate irregular and meaningful tampered regions and multiply duplicated regions. In addition, it can also locate the duplicated regions in digital images that are distorted by adding white Gaussian noise, Gaussian blurring and their mixed operations.

In speaker recognition, most of the computation originates from the likelihood computations between feature vectors of the unknown speaker and the models in the database. In this paper, we concentrate on optimizing Mel Frequency Cepstral Coefficient (MFCC) for feature extraction and Vector Quantization (VQ) for feature modeling. We reduce the number of feature vectors by pre-quantizing the test sequence prior to matching, and number of speakers by ruling out unlikely speakers during recognition process. The two important parameters, Recognition rate and minimized Average Distance between the samples, depends on the codebook size and the number of cepstral coefficients. We find, that this approach yields significant performance when the changes are made in the number of mfcc's and the codebook size. Recognition rate is found to reach upto 89% and the distortion reduced upto 69%.

State of the art algorithms for people or vehicle detection should not only be accurate in terms of detection performance and low false alarm rate, but also fast enough for real time applications. Accurate algorithms are usually very complex and tend to have a lot of calculated features to be used or parameters available for adjustments. So one big goal is to decrease the amount of necessary features used for object detection while increasing the speed of the algorithm and overall performance by finding an optimum set of classifier variables. In this paper we describe algorithms for feature selection, parameter optimisation and pattern matching especially for the task of pedestrian detection based on Histograms of Oriented Gradients and Support Vector Machine classifiers. Shape features were derived with the Histogram of Oriented Gradients algorithm which resulted in a feature vector of 6318 elements. To decrease computation time to an acceptable limit for real-time detection we reduced the full feature vector to sizes of 1000, 500, 300, 200, and 160 elements with a genetic feature selection method. With the remaining features a Support Vector Machine classifier was build and its classification parameters further optimized to result in less support vectors for further improvements in processing speed. This paper compares the classification performance, of the different SVM's on real videos (some sample images), visualizes the chosen features (which histogram bins on which location in the image search feature) and analyses the performance of the final system with respect to execution time and frame rate.

Content-based search and retrieval of video data has become a challenging and important issue. Different methods[1,2,3] have been proposed to develop video retrievals systems to achieve better performance in terms of accuracy. The proposed technique uses wavelettransforms that areCosine Wavelet Transform, Walsh, Kekre, Slant and Hartley Wavelet Transform. The benefit of energy compaction of fewer transformedcoefficients is taken to reduce the feature vector size by taking fractional coefficients[5]of transformed frames of video. Smaller feature vector size results in less time for comparison of feature vectors resulting in faster retrieval of images. The feature vectors are extracted and coefficients sets are considered as feature vectors (100%, 6.25%, 3.125%, 1.5625%, 0.7813%, 0.39%, 0.195%, 0.097%, 0.048%, 0.024%, 0.012%, 0.006% and 0.003% of complete transformed coefficients).The database consists of 500 videos spread across 10 categories.Performance using wavelets is compared with respect to their transforms. In all considered wavelet transform, experimental result show the performance improvement in fractional coefficient of wavelet transform in content based video retrieval as compared to 100% coefficient considered.

The paper presents selected numerical data dimensionality reduction techniques and their application to reduce the size of feature vectors used in an exemplary General Shape Analysis (GSA) task. The usability of applying reduced feature vectors was experimentally tested using three Fourier transform-based shape descriptors and three data reduction approaches. The aim of the experiments was to investigate which data reduction approach is the best, i.e., gives the highest percentage effectiveness value while maintaining a minimal size of the feature vector.

Segmentation is a low-level processing aimed at the partition of an image in disjoint regions, each one homogeneous with respect to some properties like intensity, texture, shape, etc. Such a task is needed in many high-level processing and applications in such diverse fields as remote-sensing, medical imaging, video coding, and industrial automation, just to name a few. In this work of thesis, we first deal with the problem of color based image segmentation, that is based on some form of homogeneity in the color/spectral properties of the image, and then move to the more complex task of texture based segmentation, where the aim is to recognize complex structures in the image which are typically non homogeneous in terms of spectral properties. Concerning color based segmentation, we focus on a specific class of Markov Random Field (MRF) probability models, namely the Tree-Structured Markov Random Fields (TS-MRF), recently proposed to provide a hierarchical multiscale description of images. Based on such a model, the unsupervised image segmentation is carried out by means of a sequence of nested class splits, where each class is modeled as a local binary MRF. As a first contribution, we first extend the model to generic tree structures removing the unnecessary binary constraint, and then devise a new TS-MRF unsupervised segmentation technique which improves upon the original algorithm by selecting a better tree structure and eliminating spurious classes. Such results are obtained by using the Mean-Shift procedure to estimate the number of pdf modes at each node, and to obtain a more reliable initial clustering for subsequent MRF optimization. To this end, we devise a new reliable and fast clustering algorithm based on the Mean-Shift technique, which makes use of a variable-bandwidth strategy based on the k-Nearest Neighbors (k-NN) technique, and is implemented with a speed-up strategy that cuts significantly the computational complexity. Experimental results on synthetic data and real life images from the remote sensing domain prove the potential of the proposed method. Turning to texture based segmentation, we present a novel multiscale texture model, and a related algorithm for the unsupervised segmentation of color images. Elementary textures are characterized by their spatial interactions with neighboring regions along selected directions. Such interactions are modeled in turn by means of a set of Markov chains, one for each direction, whose parameters are collected in a feature vector that synthetically describes the texture. Based on the feature vectors, the texture are then recursively merged, giving rise to larger and more complex textures, which appear at different scales of observation: accordingly, the model is named Hierarchical Multiple Markov Chain (H-MMC). The Texture Fragmentation and Reconstruction (TFR) algorithm, addresses the unsupervised segmentation problem based on the H-MMC model. The ``fragmentation'' step allows one to find the elementary textures of the model, while the ``reconstruction'' step defines the hierarchical image segmentation based on a probabilistic measure (texture score) which takes into account both region scale and inter-region interactions. The performance of the proposed method was assessed through the Prague segmentation benchmark, based on mosaics of real natural textures, and also tested on real-world natural images from the Berkeley dataset. Finally, an application of the TFR algorithm for the unsupervised segmentation of multiresolution remotely sensed images is proposed, based on the significant presence of textural content in last generation high-resolution satellite images. The obtained results further assess the potential of the proposed method for such real life applications.

Face recognition can be installed in a surveillance system so that it can be used for monitoring, tracking and access control. An excellent, intelligent surveillance system should be sensitive to the objects far away from the camera. Unfortunately, due to the long-distance, objects like human faces captured by the camera are too small to identify. As to enhance the subtle color differences in the face image, in this paper we first improve the resolution of the captured image using deep convolution neural networks (DCNNs). Then the efficient features are extracted and used to do classification. As for verifying the effectiveness of the proposed method, we used three databases including AR face database, Georgia Tech face database (GT) database, and Labelled Faces in the Wild (LFW) database, altogether, to conduct the training and testing. Compared to the existing approaches, experimental results show that the identification accuracy of the proposed method outperforms any existing approaches.

Thousands of images are generated everyday, which implies the need to build an easy, faster, automated classifier to classify and organize these images. Classification means selecting an appropriate class for a given image from a set of pre-defined classes. The main objective of this work is to explore feature vector generation using Walsh transform for classification. In the first method, we applied Walsh transform on the columns of an image to generate feature vectors. In second method, Walsh wavelet matrix is used for feature vector generation. In third method we proposed to apply vector quantization (VQ) on feature vectors generated by earlier methods. It gives better accuracy, fast computation and less storage space as compared with the earlier methods. Nearest neighbor and nearest mean classification algorithms are used to classify input test image. Image database used for the experimentation contains 2000 images. All these methods generate large number of outputs for single test image by considering four similarity measures, six sizes of feature vector, two ways of classification, four VQ techniques, three sizes of codebook, and five combinations of wavelet transform matrix generation. We observed improvement in accuracy from 63.22% to 74% (55% training data) through the series of techniques.

Host based Intrusion Detection Systems (HIDSs) analyze the sequence of system calls in the trace and audit log files to identify intrusive system processes. HIDSs use the frequency analysis of n-gram terms in the system call traces to identify intrusive processes. However, they are computation and resource intensive as they need to analyze a large number of n-gram input features to differentiate between normal and intrusive system processes. This put a severe limitation on their real time application while analyzing voluminous system call traces. To address this issue, we propose a computation efficient HIDS framework that initially transforms the system call traces to n-gram vector representational model and then uses a dimensionality reduction process to reduce the size of the input feature vectors. The dimensionality reduced n-gram feature vectors are finally analyzed by various machine learning based classifier models to identify intrusive processes. Performance evaluation of the proposed HIDS framework on the benchmark Australian Defense Force Academy Linux Dataset (ADFA-LD) shows that it effectively detects intrusive system processes with high accuracy and low false positive rate, while at the same time incur a minimum computational overhead.

In this study, normal (n), benign (b), and malign (m) stomach image cells have taken from faculty of Medicine the Fırat University with Light Microscope help. Total number of stomach images are 180 which be 60 n, 60 b, and 60 m. 90 of these 180 stomach images have been used for testing purposes and 90 have used for training purposes. The histograms of oriented gradient (HOG) feature vectors have been obtained for normal, benign, and malign original stomach images. The size of these HOG feature vectors is 46900×180. High-dimensional of these HOG feature vectors is reduced to lower-dimensional with Linear Discriminant Analysis (LDA). These low-dimensional data are 180×180. These low-dimensional data are classified as normal benign and malign by artificial neural network (ANN) classification. Thus, HOG_LDA_ANN method for stomach cancer images have developed. Diagnostic accuracy of classification results with this method has found as 88.9%. According to the other methods, this result has higher accuracy result. And this result has found in a shorter time.

Color Image Retrieval Using Compacted Feature Vector with Mean-Count Tree

In any multi-script environment, handwritten script classification is an unavoidable pre-requisite before the document images are fed to their respective Optical Character Recognition (OCR) engines. Over the years, this complex pattern classification problem has been solved by researchers proposing various feature vectors mostly having large dimensions, thereby increasing the computation complexity of the whole classification model. Feature Selection (FS) can serve as an intermediate step to reduce the size of the feature vectors by restricting them only to the essential and relevant features. In the present work, we have addressed this issue by introducing a new FS algorithm, called Hybrid Swarm and Gravitation-based FS (HSGFS). This algorithm has been applied over three feature vectors introduced in the literature recently—Distance-Hough Transform (DHT), Histogram of Oriented Gradients (HOG), and Modified log-Gabor (MLG) filter Transform. Three state-of-the-art classifiers, namely, Multi-Layer Perceptron (MLP), K-Nearest Neighbour (KNN), and Support Vector Machine (SVM), are used to evaluate the optimal subset of features generated by the proposed FS model. Handwritten datasets at block, text line, and word level, consisting of officially recognized 12 Indic scripts, are prepared for experimentation. An average improvement in the range of 2–5% is achieved in the classification accuracy by utilizing only about 75–80% of the original feature vectors on all three datasets. The proposed method also shows better performance when compared to some popularly used FS models. The codes used for implementing HSGFS can be found in the following Github link: https://github.com/Ritam-Guha/HSGFS.