International Journal of Signal and Imaging Systems Engineering (17 papers in press)
An automated vision based algorithm for out of context detection in images
by R. Karthika, Latha Parameswaran
Abstract: Vehicular traffic on highways is a major concern relating to safety and security. Violation of traffic rules results in fatal incidents to a very large extent. In this work an attempt has been made to detect violation of traffic rules namely vehicles in no parking and no stopping zones. Dataset consisting of cars in these zones has been used for experimentation. The proposed algorithm used Histograms of Oriented Gradient (HOG) and Adaboost cascaded classifier for training. The traffic signs have been identified using Hough transform, Circlet transform and colour analysis. Experimental results are promising with an accuracy in the range of 90 97% with recognizing no parking and no stopping sign.
Keywords: Traffic sign; car detection; HOG; Hough transform; Circlet transform; Adaboost.
A New Model Based Approach for Tennis Court Tracking in Real Time
by Manel FARHAT, Ali KHALFALLAH, Med Salim BOUHLEL
Abstract: The detection and The tracking of the tennis court is a primordial step to analyze a tennis video at higher semantic level. In this context, a new approach for tennis court tracking in real time is proposed in this paper. Our proposed system is based on model based approach allows to compute the homography between the court detected in the scene and the court model presenting the real world coordinate, in addition we take into consideration the movement of the camera which give our system more stability. For this aim, the first step is to detect the tennis court by detecting the court line and determining some interest points. We check then the motion of the camera. In case of camera motion, the court is tracked by tracking the interest points using the Lucas-Kanade algorithm. After that, These points are used by a RANSAC algorithm to estimate the homography. Then, a model based correction system is applied to reduce the projection errors. However, in case of fixed camera, we need only the model based correction system. Finally, we evaluated our system to prove their efficiency in term of speed and accuracy.
Keywords: Tracking; Ransac; Model fitting; Homography; Lucas-Kanade tracker.
Comparative Analysis of Two Leading Evolutionary Intelligence Approaches for Multilevel Thresholding
by Zhengmao Ye
Abstract: The rapid advance of artificial intelligence has made complex image processing in real time possible, especially with the help of evolutionary computation for global optimization. Bi-level thresholding is unable to produce satisfactory segmentation results for the digital images with complex object and background information. Multilevel thresholding instead has become a feasible and critical way for complex image segmentation, even in the presence of poor contrast and external artifacts. Genetic Algorithms (GAs) and Particle Swarm Optimization (PSO) are broadly recognized by far to be two dominating schemes that outperform classical ones on multilevel thresholding, which group image pixels into multiple classes in terms of the intensity level of each pixel. Qualitative analysis can usually be applied to observe the superiority of GAs and PSO to all the classical approaches. However, no convincing result is reached with respect to the performance superiority between GAs and PSO. The existing segmentation practices via GAs and PSO are either examined by visual appeals exclusively, or evaluated quantitatively assuming perfect statistical distributions. In the meanwhile, for numerous cases it is necessary to expand typical multilevel thresholding of gray scale images to that of more complicated true color images. In order to make solid and thorough comparisons, in this paper, comparative analysis of two leading multilevel thresholding approaches is conducted for true color image segmentation. The information theory is also employed so that outcomes from the systematic approaches can be further analyzed using diverse quantitative performance metrics from various aspects.
Keywords: Quantitative Analysis; Multilevel Thresholding; Genetic Algorithms (GAs); Particle Swarm Optimization (PSO).
Early Detection of Parkinsons Disease through Multimodal Features using Machine Learning Approaches
by Gunjan Pahuja, T.N. Nagabhushan, Bhanu Prasad, Ravi Pushkarna
Abstract: This study is intended to establish a relation between objective biomarkers of Parkinsons Disease (PD) based on T1-weighted MRI scans and other clinical biomarkers. It shall aid doctors in identifying the onset and progression of PD among the patients. The data set of 82 healthy and 82 PD subjects have been taken from Parkinsons Progression Markers Initiative. Voxel-based morphometry has been used for feature extraction from MRI scans. These extracted features are then combined with biochemical biomarkers for dataset enrichment. Further, a genetic algorithm is applied on this dataset to remove the redundancies and to obtain an optimal set of features. Subsequently, we used Self-adaptive Resource Allocation Network (SRAN), Extreme Learning Machine (ELM) and Support Vector Machines (SVM) to classify different subjects. It is observed that SRAN classifier gave the best performance (98.17% testing accuracy) when compared with ELM and SVM. Finally, when the extracted features in the described approach are mapped to standard Monteal Neurological Institute (MNI) template, it is found that a variation of grey matter in Thalamus is responsible for PD. The obtained results corroborate the earlier findings from the literature.
Keywords: Parkinson’s disease; Magnetic Resonance Imaging (MRI); Clinical and Plasma; Proteomic Biomarkers; Optimization; Genetic Algorithms (GA); Classification; Self-adaptive resource allocation network (SRAN); Extreme Learning Machine (ELM) and Support Vector Machines (SVM).
Copy-Move Image Forgery Detection using Direct Fuzzy Transform and Ring Projection
by Mohd Dilshad Ansari, Satya Prakash Ghrera
Abstract: Cloning(copy-move) image forgery detection (CMFD) is a pure image processing method without any support of embedded security information. Fuzzy Transform (F-Transform) is a powerful tool that encompasses both classical transforms (Fourier, Laplace, DCT and Wavelet) as well as approximation technique using fuzzy IF-THEN rules studied in fuzzy modeling. Ring projection transform (RPT) for features extraction is very effective tool as it transform two dimensional data into one dimensional with a very few components which significantly reduces the computational complexity. Additionally, in order to form matching invariant to rotation. Further, RPT converts a 2-D image into a rotation-invariant illustration in the 1-D ring projection space. We propose a new and comprise scheme of fuzzy transform and ring projection transform in copy-move image forgery detection. Firstly, fuzzy transform is employed on input image to yield highly reduced dimension representation, which is splitted into fixed-size overlapping blocks. Further, ring projection transform is applied to every block for calculating their features. These feature vectors are lexicographically sorted to make identical blocks sequentially. Finally, the duplicated blocks are filtered out using correlation coefficient, and copy-move regions were detected automatically without applying postprocessing operations. Proposed algorithm is faster and efficient in terms of execution time and accuracy.automatically without applying post-processing operations. Proposed algorithm is faster and efficient in terms of execution time and accuracy.
Keywords: Copy-move image forgery detection; Ring projection transform; Fuzzy Transform; Basic Functions; Correlation co-efficient; Feature extraction.
Modelling the Nonlinear Oscillations Due to Vertical Bouncing Using a Multi-Scale Restoring Force System Identification Method
by Yuzhu Guo, Lingzhong Guo, Vitomir Racic, Shu Wang, Stephen A. Billings
Abstract: Human vertical bouncing motion is studied using a system identification method. A multi-scale mathematical model is identified directly from real experimental data to characterise the nonlinear oscillation associated with the vertical bouncing. A new method which combines the restoring force surface method and the iterative orthogonal forward regression algorithm is proposed to determine the model structure and estimate the associated parameters. Two types of sub-models are used to study the nonlinear oscillations in different scales. Results show that the model predicted outputs provide excellent predictions of the experimental data and the models are capable of reproducing the nonlinear oscillations in both time and frequency domain.
Keywords: iterative orthogonal forward regression; iOFR; restoring force surface method; multi-scale; radial basis function; hybrid model.
Fusion Features for Robust Speaker Identification
by Ines BEN FREDJ, Youssef ZOUHIR, Kais OUNI
Abstract: Speaker's identification systems aim to identify, through a set of speech parameters, the speaker‟s identity. Thus, a relevant speech representation is required. For this purpose, we suggest to combine spectral parameters as the Mel Frequency Cepstral Coefficients (MFCC) and the Perceptual Linear Predictive coefficients (PLP) and prosodic parameter such as the signal fundamental frequency (F0). There are two main classes for F0 estimation divided into temporal and spectral methods. We employ the Sawtooth Waveform Inspired Pitch Estimator (SWIPE) algorithm for F0 estimation. It is based on the pitch estimation in the frequency domain. In addition, we evaluate the Gaussian mixture model-universal background model (GMM-UBM) for the modeling purpose. Experiments are involved on Timit database. Identification rates are promising and prove the benefit of the combination for MFCC and PLP rather than using each feature separately and this mainly for noisy data.
Keywords: Clean and noisy data; Fundamental Frequency; Fusion features; GMM-UBM; Maximum Likelihood Estimation; MFCC; PLP; Speaker identification; Speech; SWIPE; Timit.
Medical Image Fusion Using PCNN and Poisson-Hidden Markov Model
by Biswajit Biswas, BIPLAB SEN
Abstract: The combination of multiple images into a single image without loss of any generosity, the process is known as image fusion. In practice, the proper fusion of multiple images into a single image that contains significant image feature content (e.g. color, edge, corner, etc.) is a difficult problem in the domain of digital image processing. In medical diagnosis, the Magnetic resonance imaging(MRI) gives the brain tissue anatomy without any functional information whereas the Positron emission tomography(PET) image gives the brain function with low spatial resolution. Whereas, the image fusion combines the spatial resolution of the functional images by merging them with a high-resolution anatomic image. This paper proposes a novel medical image fusion technique based on pulse coupled neural net (PCNN) and Poisson Hidden Markov Model (PHMM) that satisfies fusion criterion. The excellency of the proposed approach is verified by the comparison with some of the state-of-the-art techniques in terms of the quality evaluation.
Keywords: Medical image fusion; PET-MRI; shearlet; PCNN; entropy.
An alphabet reduction algorithm for lossless compression of images with sparse histograms
by Slim Chaoui, Atef Masmoudi
Abstract: In this paper, we propose a new adaptive arithmetic coding for lossless image compression applying an alphabet reduction algorithm. The proposed method works under the trade-off relation between the average entropy decrease due to the image partitioning and the resulting additional cost for the description of the block alphabet sets so as to mitigates the well-known zero-frequency problem. This last appears especially for images with sparse and locally sparse histograms. The algorithm is a reduction mechanism of the alphabet set within each block by assigning to each one an as small as possible symbol set including all the really present symbols called active symbols, instead of using the nominal alphabet set. The analytical expression of the expected gain in terms of compression efficiency when using the block active symbol sets is derived. We show experimentally that the proposed method, in conjunction with adaptive arithmetic coding order-$0$ model applied for images with sparse and locally sparse histograms, provides promising compression ratios and outperforms several state-of-the-art lossless image compression standards such as JPEG2000, JPEG-LS and CALIC.
Keywords: adaptive arithmetic coding; alphabet reduction; lossless image compression; image partitioning; sparse histogram.
Hue Preserving Colour Image Enhancement models in RGB colour space without Gamut problem
by Krishna Gopal Dhal, Sanjoy Das
Abstract: All hue preserving colour image enhancement techniques are associated with the change of colour space, such as RGB to HSI, HSV, LHS, and YUV etc. In these colour spaces intensity or saturation or both have been processed then recombined with unchanged hue component to build a hue preserving method. But all the above techniques are time-consuming for colour space conversion and also introduce gamut problem for which the values of enhanced pixels may not lie within their respective range. In this study, three hue preserving models corresponding to three colour spaces viz. HSI, HSV and YUV, have been proposed by considering only RGB and CMY colour spaces. Any gray level contrast enhancement method can be successfully employed for colour image through those models. In this paper, a novel variant of Histogram Equalization (HE) has been proposed based on entropy based segmentation to enhance the contrast. The proposed variant called Entropy Based Brightness Preserved Dynamic Histogram Equalization (EBBPDHE) is a modification of Brightness Preserved Dynamic Histogram Equalization (BPDHE).
Keywords: HSI; HSV; YUV; Hue Preserving; Gamut; Brightness Preserved; Histogram Equalization.
Speckle Noise Reduction for 3-D Ultrasound Images by Optimum Threshold Parameter Estimation of Bi-dimensional Empirical Mode Decomposition Using Fisher Discriminant Analysis Discriminant Analysis
by Rafid Mostafiz, MOHAMMAD MOTIUR RAHMAN, Mithun Kumar PK, Mohammad Ashraful Islam
Abstract: This paper presents an approach of speckle noise reduction for 3D ultrasound images using Bi-dimensional Empirical Mode Decomposition (BEMD). 3D ultrasound is a popular diagnostic system in assessing the progression of diseases for its diverse benefits and application. Speckle noise often obscures the fine details and degrades the spatial resolution and, contrast quality that makes the interpretation of ultrasound image more difficult. The proposed method estimates an optimum threshold value of Intrinsic Mode Functions (IMFs) using Fisher Discriminant Analysis (FDA) for reducing the speckles in 3D volume of ultrasound images. FDA has applied on 2D IMFs, then explored and extended to 3D. The 3D volume rendering is performed on the basis of integrating 2D slice images that provide strong speckle reduction and edge preservation. The experiment result has compared with the several other state-of-the-art threshold methods. The proposed method is also good in edge preservation and contrast resolution.
Keywords: EMD; BEMD; IMF; Ultrasound imaging; 3D volume visualization; FDA; Speckle noise; Optimum threshold.
Fast Template-Based Technique to Extract Optic Disc from Colored Fundus Images Based on Histogram Features
by Baydaa Al-Hamadani
Abstract: The process of localizing and extracting optic disc (OD) from colored fundus images is of much benefit to the process of diagnosing several eye diseases. This paper presents a fast and robust approach to extract OD by employing the histogram features of the input image and matching it with the histogram of a template image that has no pathological features. These steps result in an image with its own OD region and other pathological structures have more color intensity than the original image. The proposed method locate part of the OD region first and then expands it to include all the required region base on morphological OD features such as location, area, and color intensity. The testing results against 1540 fundus images taken from five public databases show that the proposed technique succeeded in extracting OD region from challenging images and it achieved 97.66%, 96.93%, 99.7% for sensitivity, specificity, and accuracy respectively with a competitive average execution time equal to 2.5s.
Keywords: Optic Disk (OD); fundus images; histogram matching; morphological features; speed; accuracy.
Nakagami-m Channel Secrecy: A receiver assisted scheme
by Olawoyin Lukman
Abstract: this paper considers an end-user assisted jamming scheme to achieve a secure transmission of information over Nakagami-m} fading channel in wireless broadcast system. The use of full-duplex (FD) antenna at the legitimate receiver (LR) was introduced in such a way that the LR will contribute additional noise to the system to degrade the eavesdropper's channel. This approach shows that secure transmission is achievable and there is a considerably secrecy improvement when compared with the conventional use of artificial noise (AN) scheme. Analysis and simulation results show an improvement in the secrecy performance metrics.
Keywords: Artificial noise; full-duplex; eavesdropper; physical layer; secrecy outage probability; secrecy capacity.
Comparison of Rotation Invariant Local Frequency, LBP and SFTA methods for Breast abnormality classification
by Spandana Paramkusham, Kunda M.M.Rao, B.V.V.S.N Prabhakar Rao
Abstract: Breast cancer is the second most prominent cancer diagnosed among women. Many studies demonstrate that early detection and diagnosis is the key to decrease fatality rate. Digital mammography is one of the effective imaging modalities used to detect breast cancer in early stages. Computer aided detection systems help radiologists to detect and diagnose abnormalities earlier and faster in mammogram. In this paper, a comprehensive study is carried out on different feature extraction methods for classification of abnormal areas in mammogram. The prominent techniques used for feature extraction in this study are local binary pattern (LBP), rotation invariant local frequency (RILF) and segmented fractal texture analysis (SFTA). Features extracted from these techniques are then fed to a support vector machine (SVM) classifier for further classification via ten-fold cross validation method. The robustness of the methods under study is verified by considering 932 normal, 1157 mass and 688 microcalcification regions of interest (ROIs) of mammogram. The evaluation is performed using IRMA database for feature extraction. Experimental results have shown that the RILF method has achieved accuracy of 91.11% and 93.53% for microcalcification/normal and mass/normal classification respectively. Our statistical analysis shows that the RILF technique outperforms the LBP and SFTA techniques.
Keywords: Breast cancer; mammograms; masses; microcalcification; Feature extraction; SVM.
IF-RD Optimization for Bandwidth Compression in Video HEVC and Congestion control in Wireless networks using Dolphin echolocation optimization with FEC
by Sunil Kumar B.S.
Abstract: At present, the latest standard for coding the videos is High-Efficiency Video Coding (HEVC) technique. Owing to limitations of wireless channels on time variability and bandwidth, video transmission to multiple mobile users is difficult. Inter Frame-Rate Distortion (IF-RD) is the technique used in this paper, and it provides more clear videos with a reduction in motion signal by compressing bandwidth and enhancing optimization in HEVC. First, the multi-resolution encoder encodes the video by employing a fusion strategy called IF-RD optimization strategy. In order to improve performance and achieve noiseless transmission, the Forward Error Correction (FEC) and dolphin echolocation optimization techniques are integrated. Token-Based Congestion control (TBCC) reduces messages that are dropped due to congestion in the network. It also provides computation time. By adopting HEVC, the system performs well and achieves better compression in bandwidth and saves better Peak signal-to-noise ratio (PSNR) data rate with no extra time for encoding. Finally, the experimental results revealed that the proposed method achieves less computation (i.e.,) 80.08% better than the conventional methods.
Keywords: Video Coding; HEVC; Inter Frame-Rate Distortion (IF-RD) Optimization; Multi-Resolution Encoding; Structural similarity index; Congestion control.
Evaluating Compressive Sensing Algorithms in Through-the-Wall Radar via F1-Score
by Ali Muqaibel, Ali ALBELADI
Abstract: To achieve high resolution Through-the-Wall Radar Imaging (TWRI), long wideband antenna arrays need to be considered, thus resulting in massive amounts of data. Compressive Sensing (CS) techniques resolve this issue by allowing image reconstruction using much fewer measurements. The performance of different CS algorithms, when applied to TWRI, has not been investigated in a comprehensive and comparative manner. In this paper, popular CS algorithms are evaluated, to see which are most suitable for TWRI applications. As for the evaluation criteria, the notion of F_1-score is adopted and used in the context of TWRI; thus emphasizing the algorithms ability to reconstruct an image with correctly detected targets. Algorithms responses to different levels of SNR and compression rate are evaluated. Numerical results show that for systems with low SNR, alternating direction based algorithms work better than others. When the SNR is high, algorithms depending on spectral gradient-projection methods give good results even with high compression rates.
Keywords: Through-the-wall radar imaging; compressive sensing; F1-score.
Analysis and Estimation of Traffic Density: An Efficient Real Time Approach using Image Processing
by Shreekanth Thotappa, Madhukumar M
Abstract: Nowadays traffic density is very high in most of the urban areas, because of the increase in the number of vehicles. Traffic congestion is a very common problem that leads to more lay-out time in traffic. In order to address this issue, an algorithm has been proposed in this work for traffic flow monitoring and analysis in real time based on image processing techniques. This paper describes a method of real time area and frame based traffic density estimation using edge detection for intelligent traffic control system. Area occupied by the edges of vehicles will be considered to estimate traffic density. The system will automatically estimate the traffic density of each road by calculating the area occupied by traffic which in turn will help to determine the duration of each traffic light. The main role of this study lies in the development of a new technique that detects traffic density according to the area occupied by the edges of vehicles for controlling traffic congestion. The proposed algorithm was evaluated on a 30sec video dataset which was sampled into 8 frames and yielded an average accuracy of 98.07%. This is comparable with the existing algorithms in the literature.
Keywords: Image processing; Image Cropping; Canny Edge detection; Traffic Velocity; Traffic density; intelligent traffic control.