Search

Today’s hardware platforms have parallel processing capabilities and many parallel programming models have been developed. It is necessary to research an efficient implementation of compute-intensive applications using available platforms. Dense matrix-matrix multiplication is an important kernel that is used in many applications, while it is computationally intensive, especially for large matrix sizes. To improve the performance of this kernel, we implement it on the graphics processing unit (GPU) platform using the tiling technique with different tile sizes. Our experimental results show the tiling approach improves speed by 56.89% (2.32× faster) against straightforward (STF). And tile size of 32 has the highest speed compared to other tile sizes of 8 and 16.

Today’s hardware platforms have parallel processing capabilities and many parallel programming models have been developed. It is necessary to research an efficient implementation of compute-intensive applications using available platforms. Matrix-matrix multiplication is an important kernel used in many scientific and multimedia applications. This kernel is a compute-intensive program too. This paper aims to efficiently implement this kernel using different parallel programming models such as SIMD, OpenMP, and OpenCL. Our experiment results for different matrix sizes show that speedups up to 6.5x, 3.2x, 16.7x, and 32x are achieved for SIMD, OpenMP, hybrid OpenMP-SIMD, and OpenCL implementations over-optimized serial implementation, respectively.

Automatic facial expression recognition, which has many applications such as drivers, patients, and criminals' emotions recognition, is a challenging task. This is due to the variety of individuals and facial expression variability in different conditions, for instance, gender, race, colour and changing illumination. In addition, there are many regions in a face image such as forehead, mouth, eyes, eyebrows, nose, cheeks and chin, and extracting features of all these regions are expensive in terms of computational time. Each of the six basic emotions of anger, disgust, fear, happiness, sadness and surprise affect some regions more than the other regions. The goal of this study is to evaluate the performance of enhanced local binary pattern, pyramid histogram of oriented gradients feature‐extraction algorithms and their combination in terms of recognition accuracy, feature vector length and computational time on one, two and three combined regions of a face image. Our experimental results show that the combination of both feature‐extraction algorithms yields an average recognition accuracy of 95.33% using three regions, that is, the mouth, nose and eyes on Cohn–Kanade dataset. Besides, the mouth region is the most important part in terms of accuracy in comparison to eyes, nose and combination of both eyes and nose regions.

Automatic analysis, understanding typical activities, and identifying vehicle behaviour in crowded traffic scenes are fundamental and challenging tasks for traffic video surveillance. Some recent researches have been using machine learning approaches to extract meaningful patterns occurring in a traffic scene, for example, intersection. In this regard, we convert visual patterns and features to visual words using dense and sparse optical flow and learning traffic motion patterns with group sparse topical coding (GSTC) algorithm. In the first step of the proposed algorithm, the input traffic video is divided into non‐overlapping clips. After that, motion vectors are extracted using dual TV‐L1 as a dense optical flow and Lucas–Kanade as a sparse optical flow and converted to flow words. For learning traffic motion patterns, the GSTC algorithm, that is, a non‐probabilistic topic model (TM) has been applied. These patterns represent priors on observable motion, which can be utilised to describe a scene and answer behaviour questions such as what are the motion patterns in a traffic scene and what is going on. The experimental results which have been obtained using a real dataset, QUML, show that the combination of the GSTC + dual TV‐L1 extracts more traffic motion patterns in comparison with the GSTC + Lucas–Kanade and previous studies.

The definition of an image's category from a database with huge texture categories needs massive computation and time cost. Existing texture classification works focus on texture representation to improve the accuracy and efficiency of classification. This research wants to reduce the categories of the main classifier to decrease the comparison time of classification. To overcome computation time, a features' value range (FR) approach to enhance the throughput of texture classification is proposed. The proposed approach decreases the number of candidate categories as a pre‐classifier in a two‐step serial classification. With the decrease in the number of candidates, the main classifier can work on a few categories to find the final category. Here, configuration parameters are defined and some criteria are proposed for evaluating the FR approach. The performance of the FR is evaluated in the presence of different levels of Gaussian noise. Finally, it is shown that using effective features (EF) and hardware implementation approaches can extend the applicability of the FR approach. Experimental results depicted that the throughput of the final decision increased up to 14.85× with considerable reliability.

Today, manipulating, storing, and sending digital images are simple and easy because of the development of digital imaging devices from hardware and software points of view. Digital images are used in different contexts of people’s lives such as news, forensics, and so on. Therefore, the reliability of received images is a question that often occupies the viewer’s mind and the authenticity of digital images is increasingly important. Detecting a forged image as a genuine one as well as detecting a genuine image as a forged one can sometimes have irreparable consequences. For example, an image that is available from the scene of a crime can lead to a wrong decision if it is detected incorrectly. In this paper, we propose a combination method to improve the accuracy of copy–move forgery detection (CMFD) reducing the false positive rate (FPR) based on texture attributes. The proposed method uses a combination of the scale-invariant feature transform (SIFT) and local binary pattern (LBP). Consideration of texture features around the keypoints detected by the SIFT algorithm can be effective to reduce the incorrect matches and improve the accuracy of CMFD. In addition, to find more and better keypoints some pre-processing methods have been proposed. This study was evaluated on the COVERAGE, GRIP, and MICC-F220 databases. Experimental results show that the proposed method without clustering or segmentation and only with simple matching operations, has been able to earn the true positive rates of 98.75%, 95.45%, and 87% on the GRIP, MICC-F220, and COVERAGE datasets, respectively. Also, the proposed method, with FPRs from 17.75% to 3.75% on the GRIP dataset, has been able to achieve the best results.

Copying and pasting a patch of an image to hide or exaggerate something in a digital image is known as a copy-move forgery. Copy-move forgery detection (CMFD) is hard to detect because the copied part image from a scene has similar properties with the other parts of the image in terms of texture, light illumination, and objective. The CMFD is still a challenging issue in some attacks such as rotation, scaling, blurring, and noise. In this paper, an approach using the convolutional neural network (CNN) and k-mean clustering is for CMFD. To identify cloned parts candidates, a patch of an image is extracted using corner detection. Next, similar patches are detected using a pre-trained network inspired by the Siamese network. If two similar patches are not evidence of the CMFD, the post-process is performed using k-means clustering. Experimental analyses are done on MICC-F2000, MICC-F600, and MICC-F8 databases. The results showed that using the proposed algorithm we can receive a 94.13% and 96.98% precision and F1 score, respectively, which are the highest among all state-of-the-art algorithms.

Providing huge amounts of data is a fundamental demand when dealing with Deep Neural Networks (DNNs). Employing these algorithms to solve computer vision problems resulted in the advent of various image datasets to feed the most common visual imagery deep structures, known as Convolutional Neural Networks (CNNs). In this regard, some datasets can be found that contain hundreds or even thousands of images for license plate detection and optical character recognition purposes. However, no publicly available image dataset provides such data for the recognition of Farsi characters used in car license plates. The gap has to be filled due to the numerous advantages of developing accurate deep learning-based systems for law enforcement and surveillance purposes. This paper introduces a large-scale dataset that includes images of numbers and characters used in Iranian car license plates. The dataset, named Iranis, contains more than 83,000 images of Farsi numbers and letters collected from real-world license plate images captured by various cameras. The variety of instances in terms of camera shooting angle, illumination, resolution, and contrast make the dataset a proper choice for training DNNs. Dataset images are manually annotated for object detection and image classification. Finally, and to build a baseline for Farsi character recognition, the paper provides a performance analysis using a YOLO v.3 object detector.

Coarseness is an effective description for texture analysis and many approaches capture this property of texture in different methods. In this research, we propose a new texture descriptor (CFS) that works based on coarseness and the fineness similarity score. For tuning of its configuration, we collect coarse and fine textures based on human visual perception at first. After that, we relabel the "coarse" and "fine" categories of the gathered textures during the configuration of the operator in a proposed framework. We concatenated the features of two pyramid representations with the CFS information of the original texture (PCFS). In addition, we combine the PCFS information with our last proposed feature selection approach to improve the efficiency of the CFS. We evaluated the proposed feature extraction method with classification of one of the well-known data set, Outex. Experimental results depict satisfactory performance of the CFS with very low-dimensional feature length.

This study seeks to present an interactive tool, which exploits the GUI related abilities in MATLAB, to investigate power electronic rectifiers operation. The present paper aims to show a flexible and extendable environment for steady state simulation of ideal controlled, uncontrolled, single-phase and three-phase power electronic rectifiers. This is accomplished at the presence of R, L and E loads with or without a flywheel diode. The easy application of our tool makes it feasible to be used by the teacher in the classroom. Also its short running time and the PSpice output netlist make it a remarkable alternative to the MATLAB PowerSim toolbox and PSIM software for studying ac-dc converters. Furthermore, in this paper some basic equations were introduced to analyze single and three-phase rectifiers. Because these equations are common between ac-dc converters, analyzing the different types of rectifiers will be easy for the students. The penultimate issue refers that, the proposed program can cause a variety of single and three phase controlled rectifiers accompanied by various combinations of R, L and E loads to be simulated. Ultimately, obtained results are compared to another well-known simulator’s such as PSpice to verify their accuracy.

In this paper, a controller has been presented by the root locus method based on the state space average model of the boost switching regulator with all of the converter’s parameters and uncertainties. In this model, the load current is unknown and the inductor, capacitor, diode and active switch are non ideal and have an on-state resistance. Furthermore, an on-state voltage drop has been considered for diode and active switch. By neglecting the load current and assuming the ideal elements the simplified model of the regulator has been caddied out. By these complete and simplified models, a step by step method has been proposed to design a single input single output (SISO), second order controller based on roots locus method. In this regard the controller's electronic circuit has been introduced by operational amplifiers. At the end, by simulation of the complete closed-loop system in MATLAB Simulink environment and comparing its results by the results of the regulator and controller circuits in PLECS, the accuracy of the designed controller performance has been shown.

The most popular second-order statistical texture features are derived from the co-occurrence matrix, which has been proposed by Haralick. However, the computation of both matrix and extracting texture features are very time consuming. In order to improve the performance of co-occurrence matrices and texture feature extraction algorithms, we propose an architecture on FPGA platform. In the proposed architecture, first, the co-occurrence matrix is computed then all thirteen texture features are calculated in parallel using computed co-occurrence matrix. We have implemented the proposed architecture on Virtex 5 fx130T-3 FPGA device. Our experimental results show that a speedup of 421[× yields over a software implementation on Intel Core i7 2.0 GHz processor. In order to improve much more performance on textures, we have reduced the computation of 13 texture features to 3 texture features using ranking of Haralick's features. The performance improvement is 484×.