International Journal of Computational Vision and Robotics (33 papers in press)
A New Method for 3-Dimensional Magnetic Resonance Images Denoising
by Feriel Romdhane, Faouzi Benzarti, Hamid Amiri
Abstract: Removing noise in Magnetic Resonance Images (MRI) is a crucial issue in the field of medical image processing. These images are infected by Rician noise which is a non additive noise, allows to reduce the image contrast and causes random fluctuations. Our paper proposed a new method for 3D MRI denoising based on new combination between Non-local Means filter and the Diffusion Tensor with adaptative MAD estimator Rician noise. The performance of our proposed algorithm was evaluated with respect to different quantitative measures, compared to other denoising methods which illustrate that our proposed denoising algorithm efficiently removes noise and preserves more details.
Keywords: 3D MRI; 3D denoising method; Non-Local Mean filter; Diffusion Tensor; Rician noise.
A Novel Incremental Topological Mapping Using Global Visual Features
by Nabila Zrira, El Houssine Bouyakhf
Abstract: Mapping is fundamental in the navigation task of autonomous mobile robots. In appearance based mapping, the process of detecting visual loop closing determines whether the current observation comes from a previously visited location or a new one. The purpose of this paper is to present a new method of exploring indoor environments by an autonomous mobile robot, as well as building topological maps based on global visual attributes. This method takes advantage of the small size of the GIST descriptors, and the ease of their calculation. We also make use of omnidirectional images to build a single global visual descriptor showing an entire room. Furthermore, in order to handle the problem of a visual loop closing, we have employed a formula that correctly assigns each global descriptor to its location.
Keywords: Topological mapping, Gist descriptor, visual loop closing,omnidirectional images, navigation, mobile robots, indoor environments.
Car Manufacturer and Model Recognition based on Scale Invariant Feature Transform
by Yongbin Gao, Hyo Jong Lee
Abstract: Vehicle analysis involves lisence plate recognition, vehicle type recognition, and car manufacturer and model recognition. Car manufacturer and model recognition plays an important role to provide complementary information for lisence plate recognition for unique identification of a car. In this paper, we propose a framework to recognition car manufacturer and its model based on Scale Invariant Feature Transform (SIFT). We first detect the moving car using frame differences; the resultant binary image is used to detect the frontal view of a car by a symmetry filter. The detected frontal view is used to identify a car based on SIFT algorithm. Experiment results show that our proposed framework achieves favorable recognition accuracy.
Keywords: Moving car detection; Car model recognition; SIFT
Image compression based multiple description transform coding using NSCT and OMP approximation.
by Amina NAIMI, Kamel BELLOULATA
Abstract: In this paper we present a novel Multiple Description Transform Image Coding architecture, which uses an attractive transform called Nonsubsampled Contourlet Transform NSCT. The proposed scheme combines NSCT and orthogonal matching pursuit algorithm OMP to give a sparse representation of images, aiming at solving the compression problem due to the redundancy property of NSCT. Apply a redundant transform in coding is a challenge. In this way, the well-known algorithm OMP turns to give a solution to remove the redundancies of NSCT subbands. We evaluate the performance of our multiple description image coder in the case of four descriptions that are dispatched over different channels. The experimentations show that the proposed method is efficient and the potential using NSCT than DWT in multiple description image coding, is evaluated by PSNR in each case of packet loss, where every description can reconstruct the image with acceptable fidelity, the later is much better if all descriptions are available. Simulation results have proved the efficiency of the suggested algorithm.
Keywords: Multiple description coding (MDC), Nonsubsampled contourlet transform (NSCT), Discrete wavelet transform (DWT), Orthogonal matching pursuit (OMP).
Original strategy for avoiding over-smoothing in SFS problem resolution
by Rocco Furferi, Lapo Governi, Yary Volpe, Luca Puggelli, Monica Carfagni
Abstract: With the aim of retrieving 3D surfaces starting from single shaded images, i.e. for solving the widely known shape from shading problem, an important class of methods is based on minimization techniques where the expected surface to be retrieved is supposed to be coincident with the one that minimize a properly developed functional, consisting of several contributions. Despite several different contributes can be explored to define a functional, the so called smoothness constraint is a cornerstone since it is the most relevant contribute to guide the convergence of the minimization process towards a more accurate solution. Unfortunately, in case input shaded image is characterized by areas where actual brightness changes rapidly, such a constraint introduces an undesired over-smoothing effect for the retrieved surface. The present work proposes an original strategy for avoiding such a typical over-smoothing effect, with regards to the image regions in which this is particularly undesired such as, for instance, zones where surface details are to be preserved in the reconstruction. The proposed strategy is tested against a set of case studies and compared with other traditional SFS-based methods to prove its effectiveness.
Keywords: Shape from Shading; variational approach; 3D model; over-smoothing; minimization; smoothness constraint.
Efficient Holistic Feature Basis Learning For Pedestrian Detection
by Kyaw Kyaw Htike
Abstract: Pedestrian detection is an important research area in computer vision and Artificial Intelligence due to its potential applications in pedestrian safety, elderly monitor and care, surveillance, image retrieval and video compression. Many pedestrian detection systems have been proposed and it has been pointed out in state-of-the-art research that feature extraction is one of the significant factors in improving the performance of a pedestrian detector. Therefore, much work has focused on proposing novel feature extraction schemes to improve pedestrian detection. Moreover, most are end-to-end pedestrian detection systems, making it unclear about the contribution of classifiers in the detection pipeline. In this paper, we fill in some of this gap and focus on the classification process and propose feature basis learning for holistic high dimensional feature vectors that are common in pedestrian detection. We experimentally show that it is possible to obtain superior performance by our proposed feature basis learning algorithms even on high dimensional datasets.
Keywords: Pedestrian Detection; Feature Extraction; Classifiers; Computer Vision; Feature Learning.
Road traffic sign recognition algorithm based on computer vision
by Huiming Dai, Xin Zhang, Dacheng Yang
Abstract: As road traffic sign recognition is a crucial component for automatic driver assistance systems, and it is a key problem in computer vision as well. Therefore, in this paper, we study on the problem of road traffic sign recognition utilizing the computer vision technology. The main innovation of this paper is to propose an improved convolutional neural network, and then use it to tackle the road traffic sign recognition problem. Convolutional neural network can learn features from training data set, and a convolutional network contains alternating layers of convolution and pooling. Particularly, RGB traffic images are transformed to gray scale images, and then gray scale images are input to the improved convolutional neural network. Furthermore, the fixed layers are utilized to discover region of interests, and the learnable layers are used to extract features. In general, output information of the proposed two learnable layers are input to the classifier separately, and parameters of learnable layers and the classifier are trained at the same time. Finally, GTSDB data set is chosen to make performance evaluation, among which 600 images and 300 images are regarded as training and testing data set respectively. Experimental results demonstrate that the improved CNN based traffic sign recognition performs better than the traditional CNN.
Keywords: Road traffic sign; Object recognition; Computer vision; Convolutional neural network.
Obstacle Detection System Based on Colour Segmentation Using Monocular Vision for an Unmanned Ground Vehicle
by Auday Al-Mayyahi, Weiji Wang, Phil Birch, Alaa Hussien
Abstract: A vision system-based obstacle detection system for an autonomous ground vehicle An obstacle detection system based on vision approach is introduced for an indoor unmanned ground vehicle (UGV). Coloured and solid obstacles were placed randomly in an indoor field as obstacles. These obstacles are then captured in an image by using a monocular vision to develop an obstacle detection algorithm. The obstacles are detected by analysing and processing the captured images using computer vision and image processing techniques. A camera calibration is conducted to determine the relative position and orientation of the UGV with respect to the obstacles. The camera calibration was used to find the intrinsic and extrinsic matrices. These two matrices are then combined and used to produce the perspective projection matrix. Based on the calibration process, the relative position and the offset distance in addition to the steering angle of the UGV, from the obstacles, were derived. The field geometry was used to obtain a mapped environment in the coordinates world. In this paper, a proposed algorithm was accomplished to identify the existence of the obstacles in the field, using bounding boxes around the detected obstacle. That allows the determination of the obstacles locations in a pixel coordinate frame. Thus, the depth perception was determined by using the pixel coordinates and the camera projection matrix. In this work, real-time experiments in an indoor environment are carried out using a four wheeled UGV system to demonstrate the validity and efficiency of the proposed algorithm. The outcome shows that the actual distances between the camera and the obstacles can be obtained using this technique.
Keywords: Unmanned Ground Vehicle (UGV); Obstacle Detection; Colour Segmentation; Depth Perception; Camera Calibration.
APPLICATIONS OF HYPERSPECTRAL AND OPTICAL SCATTERING IMAGING TECHNIQUE IN THE DETECTION OF FOOD MICROORGANISM
by Xu Jing, Ma Long, Wu Jie, Xu Xiaomeng, Sun Ye, Pan Leiqing, Tu Kang
Abstract: Food is very easy to contaminate microorganism during production, processing, storage and transportation, the mass propagation of microorganism can cause food deterioration so that the food-borne pollution and food poisoning will be caused, with a serious threat to human health. However, the traditional methods for microorganism detection are complicated in process, poor in timeliness or low in sensitivity and are hard to meet the increasing requirements of the rapid and accurate detection, becoming the bottleneck for food quality and safety detection. With collecting the relevant information, then algorithm processing information and finally the relevant models, the modern optical imaging technique can achieve the rapid detection of food quality information. This article reviews in detail the the latest developments of hyperspectral imaging and optical scattering techniques in the nondestructive detection of the food microbial contamination, and discusses the advantages and deficiency of the various techniques.
Keywords: hyperspectral imaging; optical scattering; food microorganism; detection.
Weighted Feature Voting Technique for Content Based Image Retrieval
by Walaa Elhady, Abdulwahab Alsammak, Shady Elmashad
Abstract: A content based image retrieval process is used to retrieve most similar images to a query from a large database of images on the basis of extracted features. Matching measures are used to find similar images by measuring how the query features are close to the features of other images in the database. In this paper, a multi features system is proposed which incorporates more than one feature in the retrieval process. The weights of these features are calculated based on the precision of each feature to reflect its importance in the retrieval process. These weights are used in a weighted feature voting technique to incorporate the role of each feature in extracting the relevant images. Also, different distance measures are used to get the highest precision of each feature. The results of applying the multi-features and multi-distances measures technique outperforms other existing methods with accuracy 86.5% for Wang database, 86.5% for UW database and 85% for Caltech101 database
Keywords: CBIR; Feature extraction; Weighted average; Matching measures; Weighted feature voting..
A vision-based non-contact area and volume estimation of irregular structures towards applications in wound measurement
by Shubhangi Shrivastava, Alex Noel Joseph Raj
Abstract: Area and volume computation of irregular structures can be done using several techniques, but accuracy and the speed of computation is become an issue. But when wounds are considered for measurements, another factor comes into picture which is the risk of contamination due to direct contact with wounds and this can lead to catastrophic situations. In order to combat such issues, there is a requirement of non-contact measurement technique and thus, the paper discusses about a vision-based non-contact method for area and volume computation of irregular structures. The methodology presented here, models the irregular structured object using Delaunay triangulation and thus computes area and volume from the model formed. The new method is experimented on certain regular and irregular shaped objects and the results obtained along with their accuracy are provided.
Keywords: irregular structures; area and volume estimation; vision-based method; non-contact method.
Edge-based singular value decomposition for full reference colour image quality assessment
by Manisha Jadhav, Yogesh H. Dandawate, Narayan Pisharoty
Abstract: Due to intense use of digital visual aids, image quality plays a crucial role in today's life. Images are subjected to degradations during image acquisition and image processing. This affects their naturalness and usefulness in different applications. Literature shows efforts are made to develop an HVS consistent image quality metric since last few decades. New image quality metrics, extension of existing image quality algorithms and their applications are being developed by researcher's community. Singular value decomposition is one of the measures which are used to quantify the amount of distortion at different distortion levels. Based on the hypothesis that the human eye is adapted to extract edge information from any natural scene, this paper presents a novel approach of introducing edge information in SVD-based image quality metric. The results are compared with SVD-based metric available in related work in literature. Proposed metric outperforms the existing metric. Also, it is extended for evaluation of colour images.
Keywords: image quality; singular value decomposition; SVD; colour model; human visual system; HVS; full reference image quality metric; edge detection.
A hybrid scheme of image compression employing wavelets and 2D-PCA
by Manoj K. Mishra, Rohit Ghosh, Susanta Mukhopadhyay
Abstract: In this paper, we have presented a method of compressing 2D grey-scale images employing wavelets and two-dimensional principal component analysis (2D-PCA). Principal component analysis (PCA) is an already established technique for image compression which primarily aims at exploiting inter pixel redundancies present in the image, while wavelet is a tool widely used in multi-resolution image processing. In the proposed method the image is subjected to a multi-resolution decomposition using wavelet. Subsequently, 2D-PCA is applied on the set of detail images at each level of resolution. The compressed form of the image is constituted by representative pairs of principal components and projection vectors from each level of resolution along with the approximate image at the coarsest resolution. The proposed method requires relatively few number of principal components (of varied dimension) to produce improved compression ratio with acceptable peak signal to noise ratio (PSNR). The method has been implemented and tested on a set of real 2D grey-scale images and the results have been assessed on both qualitative and quantitative basis by measuring parameters like compression ratio (CR), PSNR, structural similarity index measurement (SSIM) and the overall performance is found to be satisfactory.
Keywords: 2D-PCA; feature matrix; projection vector; image compression; wavelet structural similarity index.
Human action recognition by fuzzy hidden Markov model
by Jalal A. Nasiri, Nasrollah Moghadam Charkari, Kourosh Mozafari
Abstract: Hidden Markov model (HMM) has been widely applied in human action recognition. In this paper an extension of HMM called fuzzy hidden Markov model (fuzzy HMM) is used for action recognition. It tries to increase the classification performance and decrease the information loss due to feature vector quantisation. Using fuzzy concepts with HMM leads to better recognition of similar actions such as walking, jogging and running. Two feature extraction methods including skeleton and space-time approaches are used for action representation. Actions could be represented efficiently using skeleton features where scene background is plain. Space-time features are extracted directly from video, and therefore avoid possible failures of other pre-processing methods. We propose space-time-based features by considering temporal relation between them. Experimental results show the effectiveness of fuzzy HMM in human action recognition. Moreover, it is shown that fuzzy HMM leads to significant improvement in recognition of similar actions. The accuracy rates of fuzzy HMM in comparison to HMM are incremented 3.33% and 5.59% in Weizmann and KTH datasets respectively.
Keywords: human action recognition; fuzzy hidden Markov model; FHMM; skeleton features; space-time features.
A context-based algorithm for sentiment analysis
by Srishti Sharma, Shampa Chakraverty, Akhil Sharma, Jasleen Kaur
Abstract: With netizens continuing to express a range of opinions and making assessments online, it has become a challenge to mine sentiments accurately from the ever-multiplying Big Data. We present a context-driven sentiment analysis scheme with the objective of refining the degree of subjectivity during sentiment analysis. The essence of our scheme is to capture in a stable manner, the mutual influence of the sentiments of neighbouring words on the sentiment of each word in a document. A parametric influence function combines the native sentiment score of each word with the context-derived sentiment score obtained from surrounding words. We apply a genetic algorithm to fine tune the parameters of the influence function so as to obtain the best possible accuracy for a given corpus. The experimental results on hotel reviews extracted from Tripadvisor.com show an average accuracy of 73.2% which is 3.6% more than the results obtained from the baseline sentiment analysis approach using native scores obtained from SentiWordNet. Though the improvement is small, it re-affirms our belief that contextual information provides valuable reinforcement of sentiment scores especially with regard to the borderline cases where words show near neutral sentiments. We also present a comparison with alternative sentiment analysis approaches that shows the strength of our proposed context-based sentiment analysis approach.
Keywords: sentiment analysis; opinion mining; context-based sentiment analysis; SentiWordNet; word sense disambiguation; WSD.
Efficient local adaptive thresholding for night-time vehicle candidate detection
by Yeongyu Choi, Hyojin Lim, Cuong Nguyen Khac, Ju H. Park, Ho-Youl Jung
Abstract: In the current commercial automotive market, the need for intelligent headlight control systems has increased more and more. Camera-based night-time vehicle detection has become a crucial issue in determining the performance of such control systems. The purpose of this paper is to offer an answer to the question, 'Which thresholding method is suitable in terms of detection performance for a night-time vehicle candidate selection process?% For such purposes, two local adaptive thresholding methods are introduced and tested. One is local maximum-based thresholding, and the other is local mean-based thresholding. Efficient implementation methodologies are also introduced for real-time processing. Through the simulations tested on road image sequences with different exposure times, we prove that local adaptive thresholding methods have better performance than other well-known global thresholding methods. In particular, the simulations show that the proposed mean-based thresholding method has better performance on both long- and short-exposure sequences.
Keywords: adaptive thresholding; vehicle detection; computer vision.
Odia character recognition using backpropagation network with binary features
by Mamata Nayak, Ajit Kumar Nayak
Abstract: Automatic recognition of printed characters has been an area of active research since last few decades. Though considerable work has been reported for Latin, CJK, and many other popular languages, but a few works has been reported for many Indian languages. In this work, we propose a complete model with effective techniques to recognise printed offline Odia language characters. First we develop a technique to segment the image to extract characters, and then features of these segmented characters are extracted using a binary feature extraction, as well as structural feature extraction. Finally, these extracted features are used to classify character symbols using a modified backpropagation network. In each of these phases it is found that the proposed technique outperforms existing techniques and yields high accuracy rate. Further, this work has tried to include all the possible characters (approx. 2,500 characters) in training/testing unlike other works, where a small set of characters are normally considered for testing.
Keywords: optical character recognition; OCR; binary feature; Odia language; artificial neural network; ANN.
Efficient image retrievals using generalised Gaussian mixture model
by Anuradha Padala, Srinivas Yarramalle, M.H.M. Krishna Prasad
Abstract: The advancements in technology drifted the individuals towards the usage of modern devices and as a result, the usage of web services has increased exponentially. As a result, the quantity of multimedia data accessed and stored across the internet is growing swiftly. Online money transactions, purchases, tenders and many other applications, such as messaging among specific groups were made possible. The usage of social networking sites have gained popularity due to its capability of sharing the videos and audios along with messaging, however, retrieving the most relevant videos is a challenging task among these networking groups. In this paper, a concept of audio tagging is used and the relevant videos are retrieved efficiently. The generalised GMM distribution is used as classifier and MFCC features are used to identify the voices associated with the videos.
Keywords: voice tagging; retrieval; generalised GMM; MFCC; Flicker.
A new method for automatic date fruit classification
by Oussama Aiadi, Mohammed Lamine Kherfi
Abstract: Date fruit classification by human is tedious, slow and requires several workers. In this paper, we propose a method for automatic classification of dates. Because dates of the same variety may considerably vary in terms of hardness, maturity level and shape, we represent each variety with a Gaussian mixture model (GMM). Calinski-Harabasz index has been adopted to estimate the optimal number of components for each GMM. Furthermore, the normality of samples belonging to each component is checked using Mardia's multivariate tests. Our method is able to accurately classify dates in spite of the large variation within some varieties and the small variation between some varieties. Moreover, it doesn't require any human intervention. To validate our method and as, to our knowledge, no date benchmark is publicly available; we introduce a new benchmark of 5,000 images from ten varieties. Experimental results demonstrate the effectiveness and the strength of our method.
Keywords: date fruit; date classification; Gaussian mixture model; GMM; date benchmark.
Design and analysis of active vibration and stability control of a bio-inspired quadruped robot
by D. Vishal, P.V. Manivannan
Abstract: The main aim of this research work is to design and develop a bio-inspired quadruped robot for operating in an uneven terrain. The present work focus on development of a gear-train power transmission mechanism, which is more efficient, as the gear assemblies can be optimised for generating different locomotion patterns. The kinematic and dynamic analysis of the limb analysis has been carried out. The modelling of mechanism and measurement and motion analysis has been carried out using ProE Wildfire software and the results have been presented in this paper.
Keywords: bio-inspired; gear-train power transmission; locomotion pattern; kinematic and dynamic analysis; modal and harmonic analysis.
Classification improvement using an unscented Kalman filter in brain computer interface systems
by Wansu Lim, Yeon-Mo Yang
Abstract: In this paper, we propose an enhanced classification technique using an unscented Kalman filter (UKF) for brain computer interface (BCI) signal processing. Since the UKF estimates the state of a nonlinear dynamic system and parameters for nonlinear system identification, the UKF can significantly improve the performance of classification in BCI systems. As a result, we confirm the performance improvement when using the UKF in motor imagery classification in terms of accuracy, Kappa value, and confidence interval.
Keywords: brain computer interface; BCI; unscented Kalman filter; UKF; classification; statistical signal processing.
Special Issue on: Advances in Mathematical Computational Sciences
Local convergence of Cauchy-type methods under hypotheses on the first derivative
by I.K. Argyros, D. González
Abstract: We present a local convergence analysis of Cauchy-type methods free of the second derivative using hypotheses only on the first derivative. In earlier studies such as Amat et al. (2003, 2008), Hernández and Salanova (1999), Jarratt (1996), Kou (2007), Parhi and Gupta (2007), Rall (1979) and Ren et al. (2009) hypotheses up to the fourth derivative have been used to show convergence although the method requires evaluations of the function and its derivative. This way we extend the applicability of these methods. Numerical examples are provided in this study where earlier results cannot apply but the new results can apply to solve equations.
Keywords: Cauchy's method; Newton's method; local convergence.
A new fast motion estimation algorithm using adaptive size diamond pattern search with early search termination
by Shaifali Madan Arora, Navin Rajpal, Ravindra Kumar Purwar
Abstract: In this paper, a new dynamic zero motion prejudgment (ZMP) with adaptive diamond pattern search-based algorithm is suggested to enhance the search efficiency and accuracy of motion estimation (ME) in video coding. Firstly, a dynamic ZMP technique is proposed for early identification of the stationary blocks. For non-stationary blocks, a new initial search centre prediction technique is suggested. This new search centre has high probability to be near actual MV. Its distortion is compared against a dynamically predicted threshold to check if this location could be the position of actual MV. If so, search is terminated thereafter, otherwise a variable size diamond pattern is suggested to swiftly attain the global minima. Experimental results show 95% to 99% speed gain of proposed algorithm with only 0.007-0.7 dB PSNR and 0.0001-0.0073 SSIM degradation over full search. Also, the proposed algorithm shows very promising results over other fixed and dynamic search algorithms.
Keywords: fast motion estimation; early search termination; dynamic search pattern; stationary block detection; block matching.
Visually lossless coder for volumetric MRI and CT image data using wavelet transform
by B.K. Chandrika, P. Aparna, David S. Sumam
Abstract: Medical imaging modalities produce large volume of digital data each day in modern healthcare. Several techniques have been proposed for volumetric medical image data compression. In this paper, we present a novel wavelet-based visually lossless coding scheme for the compression of volumetric magnetic resonance imaging (MRI) and computed tomography (CT) images. A visual model is incorporated in the coder to identify and measure visually irrelevant information. Performance of the compression scheme is further improved by eliminating the slice redundancy. The obtained results show better compression ratio compared to results obtained with pixel-based visually lossless compression technique, without any degradation in visual quality. We compared the performance of proposed technique with standard state of the art compression codecs such as joint photographic experts group-lossless (JPEG-LS), JPEG-2000, JPEG-3D, H.264/MPEG-4 AVC, differential pulse code modulation (DPCM) and medical image lossless compression (MILC). Results show better compression ratio over that of standard lossless compression schemes without any perceivable distortion.
Keywords: visually lossless image coding; visual model; MRI and CT images; quality metrics.
Discriminative analysis of lip features for emotion recognition
by Neeru Rathee, Dinesh Ganotra
Abstract: There exist several emotion recognition techniques that use lip information in combination with other facial features. An attempt has been made in the presented work towards emotion recognition based only on lip features. Lip information is represented by extracting lip texture features and lip geometric features. Lip texture features are extracted using discrete cosine transformation while lip geometric features are extracted by modelling the lip shape using the localised active contour model. The extracted features are applied to support vector machine for emotion recognition. The proposed approach is evaluated on extended Cohn-Kanade and JAFFE database. It is evident from the experimental results that lip geometric features result in 85.76% recognition accuracy which is higher than the recognition accuracy using lip texture features 84.47%. In addition, combination of lip texture and geometric features results in 87.70% recognition accuracy, which is comparable to the state-of-the-art approaches.
Keywords: emotion recognition; facial expression recognition; active appearance model; support vector machine; SVM.
Special Issue on: Recent Advances and Emerging Topics in Computer Vision Methods and Image Analytics
Exploring Necessity and Utility of Lightweight Android Chatting Application
by Ekbal Rashid
Abstract: In this paper there is elaborate discussion about the methods and results of a field research which led to the understanding that a lightweight chatting android application would be extremely useful for people using social networking apps. The paper details the steps how the app was developed, its salient features, its use cases and finally the findings about what people thought about it once they began using it. It also discusses the usability testing and resulting iteration. The paper attempts to highlight how such an app can help in using the resources of an Android device in a better way.
Keywords: Android; application; chatting; lightweight; mobile; communication
Effective Image Retrieval Based on Hybrid Features with Weighted Similarity Measure and Query Image Classification
by Vibhav Prakash Singh, Rajeev Srivastava
Abstract: Content Based Image Retrieval (CBIR) is a wide research area in computer vision, in which unknown query image yield similar images as per the query content. An effective CBIR needs efficient extraction of low-level features, and for this many different methods have been recently proposed using colour, texture, and shape features. Most of these methods use the histogram or some variation for representing colour and other descriptors. So, all these features may require a significant amount of space and more similarity calculation. Also, the CBIR performance is not so encouraging due to gap between low-level visual features and high-level understanding. Here, an efficient CBIR system is proposed, which is based on the fusion of chromaticity-colour moments, and colour co-occurrence based small dimension features using inverse variance weighted similarity measure. In this measure, weight of a feature with high variance is low, while the weight of a feature with low variance is high. This interesting property of the varying weights, effectively retrieves relevant images. In addition, this paper also proposes a supervised query image classification and retrieval model by filtering out irrelevant class images using a multiclass support vector machine (SVM) classifier. Basically, this model categorises and recovers the category of a query image based on its visual content, and this successful categorization of images significantly enhances the performance and searching time of retrieval system. Descriptive comparative analyses confirm the effectiveness of this work. Here, we have obtained 83.83 % and 76.9 % average precision for 12 and 20 image retrieval, respectively using weighted similarity measure together with 85.6% average precision and 84.4 % recall for the query image classification framework.
Keywords: classification; chromaticity moment; colour co-occurrence; feature fusion; content based image retrieval; variance.
Facial Expression Recognition based on Eigenspaces and Principle Component Analysis
by Ashim Saha
Abstract: Facial Expression detection or Emotion Recognition is one of the risingrnfields of research on intelligent systems. Emotion plays a significant role in nonverbal communication. An efficient face and facial feature detection algorithms are required to detect emotion at that particular moment. In this work authors implemented a system that recognizes the users facial expressions from the input images, using the algorithm of Eigenspaces and Principle Component Analysis (PCA). Eigenspaces are the face images which projected onto a feature space that encodes the variation among known face images. Authors used PCA to make dimensional reduction of images in order to obtain a reduced representation of face images. The implementation is been applied on three different Facial Expressions databases, Extended Cohn-Kanade facial expression database, Japanese Female Facial Expression database and self made database in order to find out therneffectiveness of proposed method.
Keywords: Facial expression; Eigenspaces; Principle component analysis; Emotion detection; Image Processing.
Content Based Image Retrieval with Pachinko Allocation Model and a Combination of Color, Texture and Text Features
by Ahmed Boulemden, Yamina Tlili, Hamid Jalab
Abstract: Probabilistic topic models are a set of algorithms which aim to learn and discover hidden concepts responsible of generating words of documents in large archives. These models have been also used for image processing tasks such as object recognition, image annotation and image retrieval. rnWe present in this paper a content based image retrieval system (CBIR) based on pachinko allocation model (PAM) and employing a combination of color, texture and textual features. PAM has presented more efficiency compared with other topic models by the way in which it captures correlation not only between words in documents but also between different topics (concepts) responsible of their generation. Although this advantage of PAM, there is no works which explore its utility for content based image retrieval tasks and using single and multimodal image features.rnWe aim to evaluate the use of PAM for CBIRs by implementing a system based on it. We are interested also in evaluating PAM with single and multimodal image features (i.e. the use of single and combined features). In this context, PAM was applied with two different modalities of features, image global features (color and texture) and textual indexes (from associate texts with images) separately and combined.rnMean Average Precision is evaluated. The use of PAM with combination of features has slightly improved results of using it with just one modality, this opens more perspective in order to enhance results. Images from the ImageCLEF IAPR 2012 dataset have been used for experiments.rn
Keywords: Pachinko allocation;image retrieval;color moments;texture features;global feature extraction;textual modality;features combination
An automatic natural feature selection system for indoor tracking - Application to Alzheimer patient support
by Mohamed Badeche, Frederic Bousefsaf, Abdelhak Moussaoui, Mohamed Benmohammed, Alain Pruski
Abstract: In this paper, we propose an automatic selection and natural feature tracking method that uses a monocular camera for path capturing and guides the user showing him the path to be followed. The application targets Alzheimer patients for helping them in their indoor moves. By offering an automatic selection of features, the user intervention and prior knowledge of the working environment would not be required to assure the good working of the system. The general principle of the proposed method is to record the path to be followed, and then recognize it in real time using purely visual methods, using only a single camera as an acquisition sensor. The devised system could be implemented on augmented-reality glasses with one single built-in camera. The experimental results have shown that the proposed method is very promising and the application could follow accurately the required path in real time, with a satisfying robustness in a fully-contrasted and static environment.
Keywords: natural features; matching; local descriptor; optical flow; Alzheimer disease; augmented reality glasses.
Combining Zernike moment and Complex wavelet transform for Human object Classification
by Manish Khare, Om Prakash, Rajneesh Kumar Srivastava
Abstract: Human object classification is an important problem for smart video surveillance, where we classify human object in real scenes. Even though different features have been used for human object classification task, most of the existing methods adopt a single feature to classify the objects. In this paper, we have proposed a new method for human object classification, which classify the object present in a scene into one of the two classes: human and non-human. The proposed method uses combination of Daubechies complex wavelet transform and Zernike moment as a feature of object. The motivation behind using combination of these two as a features of object, because shift-invariance and better edge representation property makes Daubechies complex wavelet transform suitable for locating object as compared to real valued wavelet transform, whereas rotation invariance property of Zernike moment is also helpful for correct object identification. Therefore, combination of these two features brings about significant synthesized benefits over each single feature and other widely used features. The proposed method matches Zernike moments of Daubechies complex wavelet coefficients of objects. We have used Adaboost as a classifier for classification of the objects. The proposed method has been tested on standard dataset like INRIA person dataset. Quantitative experimental evaluation results shows that the proposed method is better than other state-of-the-art methods and gives better performance for human object classification.
Keywords: Human object classification, feature selection, Daubechies complex wavelet transform, Zernike moment, Adaboost classifier, Video Surveillance
Stairways Detection Based on Approach Evaluation and Vertical Vanishing Point
by Md. Khaliluzzaman, Kaushik Deb
Abstract: Detecting stair region and estimating the distance from a camera to stair in a stair image is the fundamental step in the implementation of autonomous stair climbing navigation, as well as alarm systems for vision impaired people. In this paper, a framework is proposed for detecting the stair region from a stair image utilizing some natural properties of a stair. One unique property of them is, every stair steps beginning and ending horizontal edge point intersects with two vertical edge points creating three connected point. These vertical edges are stair steps height and its width edge. Another property is steps of a stair appear gradually increasing order from top to bottom of a stair in a parallel arrangement. For that initially, directional Gabor filter and Canny edge detector are employed on the stair image to eliminate the influence of illumination and for detecting stair edges. Non-candidate stair edges are eliminated by performing filtering operation. Then longest horizontal edges are extracted by using a proposed edge linking method on the edge image. After that, a search method is applied for finding stair step height and its width edge point at the beginning and ending point of the longest horizontal edges. This operation is performed for detecting three connected points to validate the stair edge segments. In the next step, these validated edge segments are used to calculate the vertical vanishing point to justify the stair edges. This justification ensures that validated edge segments are arranged in an increasing parallel order from top to bottom of a stair. Finally, these increasing edge segments are verified from other stairs similar patterns. This verification is performed by utilizing the y coordinate value of the vanishing point and confirmed the detection of stair candidate region. In addition, the triangular similarity is used for distance estimation from camera to stair. The proposed framework is tested using various stair images under a variety of conditions and results are presented to demonstrate the efficiency and effectiveness.
Keywords: autonomous stair climbing navigation; alarm system; Gabor filter; illumination; vanishing point; three connected point; triangular similarity.
Special Issue on: Recent Advances in Theory and Applications of Visual Intelligence
Modeling and Simulation of the In-wheel Motor Applied in Electric Vehicle
by Shanshan Peng, Xuejiao Wang, Shipei Cheng, Rongyun Zhang
Abstract: To study the ride comfort of in-wheel motor electric vehicle, it is necessary to analyze the speed ripple and torque ripple of motor. Thus, the motion differential equation of PMSM (permanent magnet synchronous motor) is presented in this paper. Based on the vector control principle of motor and the rotor-field-orientated control, a double closed loop controlled PMSM simulation model is built through Sim Power System Toolbox. After that, the simulation analysis of the running process of motor uninfluenced and influenced by load is carried out in Matlab/Simulink. Then the performance curves of speed, torque, and stator current of motor are obtained from the simulation analysis. As last, an in-wheel motor test system is established to verify the simulation results. The research results show that the proposed simulation model of motor is reliable and can fully reflect the running status of real motor which provides basis for the further researches on the effect of motor torque ripple on the vertical vibration of vehicle suspension.
Keywords: Electric vehicle; in-wheel motor; modeling; simulation.