Статьи журнала - International Journal of Image, Graphics and Signal Processing

Все статьи: 1110

Feature Tracking and Synchronous Scene Generation with a Single Camera

Feature Tracking and Synchronous Scene Generation with a Single Camera

Zheng Chai, Takafumi Matsumaru

Статья научная

This paper shows a method of tracking feature points to update camera pose and generating a synchronous map for AR (Augmented Reality) system. Firstly we select the ORB (Oriented FAST and Rotated BRIEF) [1] detection algorithm to detect the feature points which have depth information to be markers, and we use the LK (Lucas-Kanade) optical flow [2] algorithm to track four of them. Then we compute the rotation and translation of the moving camera by relationship matrix between 2D image coordinate and 3D world coordinate, and then we update the camera pose. Last we generate the map, and we draw some AR objects on it. If the feature points are missing, we can compute the same world coordinate as the one before missing to recover tracking by using new corresponding 2D/3D feature points and camera poses at that time. There are three novelties of this study: an improved ORB detection, which can obtain depth information, a rapid update of camera pose, and tracking recovery. Referring to the PTAM (Parallel Tracking and Mapping) [3], we also divide the process into two parallel sub-processes: Detecting and Tracking (including recovery when necessary) the feature points and updating the camera pose is one thread. Generating the map and drawing some objects is another thread. This parallel method can save time for the AR system and make the process work in real-time.

Бесплатно

FeatureGAN: Combining GAN and Autoencoder for Pavement Crack Image Data Augmentations

FeatureGAN: Combining GAN and Autoencoder for Pavement Crack Image Data Augmentations

Xinkai Zhang, Bo Peng, Zaid Al-Huda, Donghai Zhai

Статья научная

In the pavement crack segmentation task, the accurate pixel-level labeling required in the fully supervised training of deep neural networks (DNN) is challenging. Although cracks often exhibit low-level image characters in terms of edges, there might be various high-level background information based on the complex pavement conditions. In practice, crack samples containing various semantic backgrounds are scarce. To overcome these problems, we propose a novel method for augmenting the training data for DNN based crack segmentation task. It employs the generative adversarial network (GAN), which utilizes a crack-free image, a crack image, and a corresponding image mask to generate a new crack image. In combination with an auto-encoder, the proposed GAN can be used to train crack segmentation networks. By creating a manual mask, no additional crack images are required to be labeled, and data augmentation and annotation are achieved simultaneously. Our experiments are conducted on two public datasets using five segmentation models of different sizes to verify the effectiveness of the proposed method. Experimental results demonstrate that the proposed method is effective for crack segmentation.

Бесплатно

Fetal Brain Planes Classification Using Deep Ensemble Transfer Learning from U-Net Segmented Fetal Neurosonography Images

Fetal Brain Planes Classification Using Deep Ensemble Transfer Learning from U-Net Segmented Fetal Neurosonography Images

Md. Nazmul Hasan, A.B.M. Aowlad Hossain

Статья научная

Fetal neurosonography is potentially used to examine the fetal brain by scanning the trans-thalamic (TT), trans-cerebellum (TC), and trans-ventricular (TV) planes. Cross-sectional analysis of these planes is useful to assess the brain anatomy, development, and abnormality for intervention and treatment plans even at the postnatal stage. To minimize the errors and processing time involved in the traditional manual subjective approach, the automatic classification of fetal brain planes is crucial. In this study, a deep learning-based method for automatically categorizing fetal brain planes from ultrasound images is proposed and evaluated. Firstly, the brain region has been segmented from the fetal brain ultrasound images using U-Net to prepare an efficient data set for the classifier model. Then, an ensemble convolutional neural network (CNN) model including well-known Inception V3, ResNet50-V2, and DenseNet-201 models with max voting is designed to classify the segmented brain planes. 2019 fetal brain ultrasound images from a widely used publicly accessible experts-annotated dataset are used to evaluate the performance of the proposed framework. The obtained results analysis shows that using the segmented images as input improves the performance of the classifier from its raw form. The gradient class activation mapping (Grad-CAM) based inspection shows noteworthy localization capability of the last convolution layer. The ensemble model has also outperformed its individual model’s performance. The suggested categorization framework is satisfactory compared to related recent works, with a testing accuracy of 97.68%. The proposed framework for fetal brain plane classification is expected to be useful for clinical applications.

Бесплатно

Fig (Ficus Carica L.) Identification Based on Mutual Information and Neural Networks

Fig (Ficus Carica L.) Identification Based on Mutual Information and Neural Networks

Ghada Kattmah, Gamil Abdel Azim

Статья научная

The process of recognition and identification of plant species is very time-consuming as it has been mainly carried out by botanists. The focus of computerized living plant's identification is on stable feature's extraction of plants. Leaf-based features are preferred over fruits, also the long period of its existence than fruits. In this preliminary study, we study and propose neural networks and Mutual information for identification of two, three Fig cultivars (Ficus Carica L.) in Syria region. The identification depends on image features of Fig tree leaves. A feature extractor is designed based on Mutual Information computation. The Neural Networks is used with two hidden layers and one output layer with 3 nodes that correspond to varieties (classes) of FIG leaves. The proposal technique is a tester on a database of 84 images leaves with 28 images for each variety (class). The result shows that our technique is promising, where the recognition rates 100%, and 92% for the training and testing respectively for the two cultivars with 100% and 90 for the three cultivars. The preliminary results obtained indicated the technical feasibility of the proposed method, which will be applied for more than 80 varieties existent in Syria.

Бесплатно

Finding Longest Common Substrings in Documents

Finding Longest Common Substrings in Documents

M.I.Khalil, M.A.Hadi

Статья научная

This paper introduces an algorithm to address the problem of finding the longest common substring between two documents. This problem is known as the longest common substring (LCS) problem. The proposed algorithm is based on the convolution between the two sequences (named major sequence (X) which is represented as array and the minor one (Y) which is represented as circular linked list. An array of linked lists is established and a new node is created for each match between two substrings. If two or more matches in different locations in string Y share the same location in string X, the corresponding nodes will construct a unique linked-list. Accordingly, by the end of processing, we obtain a group of linked-lists containing nodes arranged in certain manner representing all possible matches between sequences X and Y. The algorithm has been implemented and tested in C# language under Windows platform. The obtained results presented a very good speedups and indicated that impressive improvements had been achieved.

Бесплатно

Finger gesture detection and application using hue saturation value

Finger gesture detection and application using hue saturation value

Joy Mazumder, Laila Naznin Nahar, Md. Moin Uddin Atique

Статья научная

In this paper, we developed a mathematical model for finger gesture identification using two colored glove. The glove is designed in such a manner that wristband and middle finger of the glove are marked by blue color and other fingers are marked by red color. HSV values of those colors are implicated in order to identify red and blue colors. After detecting colors, two processes are employed for identification of fingers. One of them is the angle created at the wristband center between the middle finger and any other fingers. Other is examining the ratio between the wristband-middle finger distance and the projection of the wristband and other fingers distance on wrist-middle finger joining line. For both processes, the middle finger must present in order to identify the fingers. After identification of fingers gesture using both methods, an application of finger detection is presented here by changing a PowerPoint slide. This mathematical model was tested on several conditions and got the accuracy of more than 82%.

Бесплатно

Fingerprint Authentication in Digital Watermarking Using YCbCr Colour Space & 2D Walsh Code

Fingerprint Authentication in Digital Watermarking Using YCbCr Colour Space & 2D Walsh Code

B P Mishra, P Das, H N Pratihari

Статья научная

During the recent development in image manipulating software and vast use of Internet, it is now becomes very difficult to protect the images that are precious and need to be secured so that they will survive against several image modification attacks. This paper represents a new technique to produce robust and efficient Twin blind digital Watermarking with the use of 2-D Walsh code and Discrete Cosine transform. Authentication matching process is introduced during the extraction process to provide extra security to the Host image. Both the Watermarks are embedded into host image through Walsh Code conversion. In this technique, the Embedding and extraction of Watermark is simpler than the other transform previously used. The proposed algorithm uses the YCbCr colour elements of the colour images in DCT province with low frequency components. During the first step the Principal Watermark i.e Hand written signature is embedded through 2-D Walsh coding and then the resultant watermark i.e Biometric fingerprint is embedded to the first Watermarked image through 2-D Walsh coding. The De-watermarking is dawn by checking the Authentication through Biometric fingerprint matching method. The technique is accessed by analyzing various performance parameters like SSIM, PSNR and NC. Further, the evaluation is made through various attacks by using StirMark tool. It was observed from the result that, by utilizing 2-D Walsh coding technique, better robustness is maintained and the proposed technique survived against various attacks such as JPEG compression, median, noise etc.

Бесплатно

Fingerprint Image Fusion: A Cutting-edge Perspective on Gender Classification via Rotational Invariant Features

Fingerprint Image Fusion: A Cutting-edge Perspective on Gender Classification via Rotational Invariant Features

Shivanand Gornale, Abhijit Patil, Khang Wen Goh, Sathish Kumar, Kruthi R.

Статья научная

In this cutting-edge technological milieu, fingerprints have become an alternative expression for the biometrics system. A fingerprint is one of the perceptible biometric modals which is predominantly utilized in almost all the security, and real-life applications. Fingerprints have many inherent rotational features that are mostly utilized for person recognition besides these features can also be utilized for the person gender classification. Thus, the proposed work is a novel algorithm which identifies the gender of an individual based on the fingerprint. The image fusion and feature level fusion technique are deliberated over the fingerprints with rotational invariant features. Experiments were carried on four state-of-the-art datasets and realized promising results by outperforming earlier outcomes.

Бесплатно

Fire extinguishing system for high-rise buildings and rugged mountainous terrains utilizing quadrotor unmanned aerial vehicle

Fire extinguishing system for high-rise buildings and rugged mountainous terrains utilizing quadrotor unmanned aerial vehicle

Abdel Ilah N. Alshbatat

Статья научная

Nowadays, fighting fires in rugged mountainous terrains or in high-rise buildings are both difficult and dangerous. High-rise buildings are containing floors at such a height or position that the deployment of external firefighting equipment and rescue operations may not be feasible or practical. Meanwhile, reaching farms located in a rugged mountainous terrain with firefighting vehicles is often impossible. From this perspective, to meet the need for a fast way to extinguish fire in an area that is hard to be approached by the conventional methods, and to offer the highest level of safety for the public and firefighters; fire extinguishing system is proposed in this paper. The system is structured with six units: Quadrotor Unmanned Aerial Vehicle (QUAV), Robot (Release Mechanism), Automatic Electric_Spring Operated Gun, Fire Extinguishing Ball, Collision Avoidance System, and a Camera. Quadrotor will carry a specific payload and be capable of throwing an extinguishing balls in an area that is chosen by the operator. The proposed system has been implemented, constructed, and tested in an actual scenarios. Experimental results demonstrate the feasibility of our drone in extinguishing fire in its initial stages and of being safe to hover over a fire, drop a fire extinguishing ball, fly back to the firefighter, and hover at 2.5 meters in the air so that it can be reloaded with a new ball without losing its stability.

Бесплатно

Fixed Windows in Fractional Fourier Domain

Fixed Windows in Fractional Fourier Domain

Rahul Pachauri, Rajiv Saxena, Sanjeev N. Sharma

Статья научная

In this study, some mathematical relations have been derived for the useful parameters of fixed window functions in fractional Fourier transform (FRFT) domain. These reported expressions are also verified with the simulation studies. The FRFT provides an important extension to conventional Fourier transform with an additional degree of freedom by which these parameters of window functions can be controlled while inherent time domain behavior of the windows remains intact. The behavior of fixed windows on time-frequency plane has been varied by varying the FRFT order. The obtained variability in the window functions has been applied in the designing of FIR filters.

Бесплатно

Fluid Temperature Detection Based on its Sound with a Deep Learning Approach

Fluid Temperature Detection Based on its Sound with a Deep Learning Approach

Arshia Foroozan Yazdani, Ali Bozorgi Mehr, Iman Showkatyan, Amin Hashemi, Mohsen Kakavand

Статья научная

The present study, the main idea of which was based on one of the questions of I.P.T.2018 competition, aimed to develop a high-precision relationship between the fluid temperature and the sound produced when colliding with different surfaces, by creating a data collection tool. In fact, this paper was provided based on a traditional phenomenological project using the well-known deep neural networks, in order to achieve an acceptable accuracy in this project. In order to improve the quality of the paper, the data were analyzed in two ways: I. Using the images of data spectrogram and the known V.G.G.16 network. II. Applying the data audio signal and a convolutional neural network (C.N.N.). Finally, both methods have obtained an acceptable precision above 85%.

Бесплатно

Frame work for expression invariant face recognition system using warping technique

Frame work for expression invariant face recognition system using warping technique

Deepti Ahlawat, Vijay Nehra, Darshana Hooda

Статья научная

Facial expressions, usually has an adverse effect on the performance of a face recognition system. In this investigation, expression invariant face recognition algorithm is presented that converts input face image with an arbitrary expression into its corresponding neutral facial image. In the present study, deep learning algorithm is used to train classifiers for reference key-points, where key-points are located and deep neural network is trained to make the system able to locate the landmarks in test image. Create an intermediate triangular mesh from the test and reference image and then warp it using affine transform and take the average of the normalized faces. To extract the features presented in the result image shift invariant feature extraction technique is used. Finally, results are compared and the recognition accuracy is determined for different expressions. The present work is tested on three different databases: JAFFE, Cohn-Kanade (CK) and Yale database. Experimental results show that the expression invariant face recognition method is very robust to variety of expressions and recognition accuracy is found to be 97.8 %, 96.8% and 95.7% for CK, JAFFE and Yale databases respectively.

Бесплатно

Fruit recognition using color and morphological features fusion

Fruit recognition using color and morphological features fusion

Myint San, Mie Mie Aung, Phyu Phyu Khaing

Статья научная

It is still difficult to recognize the kind of fruit which are of different colors, shapes, and textures. This paper proposes a features fusion method to recognize five different classes of fruits that are the images from the fruit360 dataset. We are processed with four stages: pre-processing, boundary extraction, feature extractions, and classification. Pre-processing is performed to remove the noise by using the median filter, and boundary extraction are operated with the morphological operation. In feature extraction, we have extracted two types of features: color, and morphological features of the image. Color features are extracted from the RGB color channel, and morphological features are extracted from the image that detected the boundary of fruit by using morphological operations. These two types of features are combined in a single feature descriptor. These features are passed to five different classifiers: Naïve Bayes (NB), Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbor (KNN), and Random Forest (RF). In the study, the accuracy that classified with Random Forest (RF) classifier for the proposed feature fusion method is better than the other classifiers, such as Naïve Bayes (NB), Logistic Regression (LR), Support Vector Machine (SVM), Decision Tree (DT), K-Nearest Neighbor (KNN).

Бесплатно

Fuzzy Logic Based Four Step Search Algorithm for Motion Vector Estimation

Fuzzy Logic Based Four Step Search Algorithm for Motion Vector Estimation

Suvojit Acharjee, Sheli Sinha Chaudhuri

Статья научная

Visual information is very much important for human to perceive, recognize and understand the surrounding world. As we live in the age of multimedia video sequences are very useful to us for providing information. Video involves a huge amount of data. So video compression is necessary Motion compensation has lot of computation in total video compression process. Fast motion vector estimation is a key-factor in video coding standard. Full search algorithm is the best algorithm between all the block matching algorithms to estimate the motion vector estimation with a huge computation cost. The challenge is to reduce the computational complexity of Full Search algorithm without losing too much quality at the output. In this paper we propose to implement the fuzzy logic based Four Step Search algorithm which performs better than other block matching algorithms.

Бесплатно

Fuzzy Logic Based Three Step Search Algorithm for Motion Vector Estimation

Fuzzy Logic Based Three Step Search Algorithm for Motion Vector Estimation

Suvojit Acharjee, Sheli Sinha Chaudhuri

Статья научная

Motion compensation process is the most computationally expensive operation in the entire video compression process. Fast motion estimation technique plays a very important role in video compression standard. In Block Matching Algorithm Full Search Algorithm produces the best result for motion vector estimation. But Full Search algorithm is a time consuming and computationally expensive process. The Challenge is to reduce the computational complexity of Full Search algorithm without losing too much quality at the output. In this paper we propose to implement the fuzzy logic based Three Step Search algorithm. This Algorithm performs better than the Three Step Search(TSS), New Three Step Search(NTSS), Four Step Search(FSS) algorithm.

Бесплатно

Fuzzy entropy based impulse noise detection and correction method for digital images

Fuzzy entropy based impulse noise detection and correction method for digital images

S.Vijaya Kumar, C.Nagaraju

Статья научная

Impulse noise is the prime factor which reduces the quality of the digital image and it erases the important details of the images. De-noising is an indispensable task to restore the image features from the corrupted low- quality images and improve the perceptual quality of images. Several techniques are used for image quality enhancement and image restoration. In this work, an image de-noising scheme is developed to detect and correct the impulse noise from the image by using fuzzy entropy. The proposed algorithm is designed in two phases, such as noise detection phase, and correction phase. In the noise detection phase, the fuzzy entropy of pixels in a window of interest (WoI) is computed to detect whether the pixel is noisy or not. The Fuzzy entropy of pixel greater than specified alpha cut value will be considered as noise pixel and submitted to correction phase. In the correction phase noise pixel value is replaced with a fuzzy weighted mean of the un-corrupted pixels in the WoI. The proposed Fuzzy entropy based impulse noise detection and correction method are implemented using MATLAB. The experimentation has been carried out on different standard images and the analysis is performed by comparing the performance of the proposed scheme with that of the existing methods such as DBA, MDBUTMF, AMF, NAFSM, BDND, and CM , using PSNR, SSIM, and NAE as metric parameters. The proposed method will give good results compared to state of the art methods in image restoration.

Бесплатно

GLCM based Improved Mammogram Classification using Associative Classifier

GLCM based Improved Mammogram Classification using Associative Classifier

Jyoti Deshmukh, Udhav Bhosle

Статья научная

Among women, 12% possibility of developing a breast cancer and 3.5% possibility of mortality due to this cause is reported [1]. Nowadays early detection of breast cancer became very important. Mammogram - a breast X-ray is used to investigate and diagnose breast cancer. In this paper, authors propose GLCM (Grey Level Co-occurrence Matrix) feature based improved mammogram classification using an associative classifier. Mining of association rules from mammogram dataset discovers frequently occurring patterns. It depends on user specified minimum confidence and support value. This dependency causes an increase in search space. The authors propose two-phase optimization procedure to overcome these limitations. The initial phase comprises feature optimization by adopting proposed PreARM (Pre-processing step for Association Rule Mining) method. The next phase comprises association rule optimization by adopting proposed ESAR (Extraction of Strong Association Rules) method to generate efficient, highly correlated and robust rules. Proposed associative classification method is substantiated by adapting authentic MIAS and DDSM mammogram database. The experimentation concedes 91% and 90% trimming of GLCM features and association rules by adopting PreARM and ESAR algorithms respectively. Using optimized association rules, the classification accuracies procured for MIAS and DDSM datasets are 92% and 94% respectively. Area under Receiver Operating Characteristic (ROC) curves obtained by proposed system for MIAS and DDSM datasets are 0.9656 and 0.9285 respectively. Results of GLCM based associative classifier are compared with GLCM based Random Forest (RF), an ensemble learning method. The experimental result shows that GLCM based associative classifier outperforms RF method with respect to accuracy and AUC, and it is a promising method for mammogram classification.

Бесплатно

GPU-Based Volume Rendering for 3D Electromagnetic Environment on Virtual Globe

GPU-Based Volume Rendering for 3D Electromagnetic Environment on Virtual Globe

Chao Yang, Lingda Wu

Статья научная

Volume rendering is an important and effect algorithm to represent 3D volumetric data and 3D visualization on electromagnetic environment (EME) is one of the most important research fields in 3D battlespace. This paper presents a novel framework on visualizing the 3D EME by direct volume rendering on virtual globe. 3D power volumetric data is calculated based on the Longley-Rice radio propagation model (Irregular Terrain Model, ITM), which takes into account the effects of irregular terrain and atmosphere, and we use GPU-accelerated method to compute the EME volumetric data. The EME data are rendered using direct volume rendering method on virtual globe by assigning different color and opacity depending on user’s interactive input with color picker. We also propose an interactive method to show detailed information of EME at given place. This approach provides excellent decision supporting and plan-aiding for users.

Бесплатно

GWAgeER – A GroupWise Age Ranking Framework for Human Age Estimation

GWAgeER – A GroupWise Age Ranking Framework for Human Age Estimation

Olufade F. W. Onifade, Damilola J. Akinyemi

Статья научная

The task of estimating the age of humans from facial image is a challenging one due to the non-linear and personalized pattern of aging differing from one individual to another. In this work, we investigated the problem of estimating the age of humans from their facial image using a GroupWise age ranking approach complemented by ageing pattern correlation learning. In our proposed GroupWise age-ranking approach, we constructed a reference image set grouped according to ages for each individual in the reference set and used this to obtain age-ranks for each age group in the reference set. The constructed reference set was used to obtain transformed LBP features called age-rank-biased LBP (arLBP) features which were used with attached age-ranks to train an age estimating function for predicting the ages of test images. Our experiments on the publicly available FG-NET dataset and a locally collected dataset (FAGE) shows the best known age estimation accuracy with MAE of 2.34 years on FG-NET using the leave-one-person-out strategy.

Бесплатно

Gait Recognition for Human Identification Using Fourier Descriptor and Anatomical Landmarks

Gait Recognition for Human Identification Using Fourier Descriptor and Anatomical Landmarks

Mridul Ghosh, Debotosh Bhattacharjee

Статья научная

This paper presents a gait recognition method which is based on spatio-temporal movement characteristics of human subject with respect to surveillance camera. Different measures, like leg rise from ground (LRFG), the angles created between the legs with the centre of Mass (ABLC), angles created between the Centre of Mass Knee and Ankle with the (CKA), angles created between Centre of Mass, Wrist and knee (CWK), the distances between the control points and centre of Mass (DCC) have been taken as different features. Fourier descriptor has been used for shape extraction of individual frames of a subject. Statistical approach has been used for recognition of individuals based on the n feature vectors, each of size 23(collected from LRFG, ABLCs, CKA, CWK and DCCs) for each video frame. It has been found that recognition result of our approach is encouraging with compared to other recent methods.

Бесплатно

Журнал