Computer Image Processing Techniques Research Articles

Dynamic Gesture Controlled User Interface Expert HCI System using Adaptative Background Masking: An Aid to Prevent Cross Infections

Human-Computer Interaction (HCI) interfaces need unambiguous instructions in the form of mouse clicks or keyboard taps from the user and thus gets complex. To simplify this monotonous task, a real-time hand gesture recognition method using computer vision, image, and video processing techniques has been proposed. Controlling infections has turned out to be the major concern of the healthcare environment. Several input devices such as keyboards, mouse, touch screens can be considered as a breeding ground for various micro pathogens and bacteria. Direct use of hands as an input device is an innovative method for providing natural HCI ensuring minimal physical contact with the devices i.e., less transmission of bacteria and thus can prevent cross infections. Convolutional Neural Network (CNN) has been used for object detection and classification. CNN architecture for 3d object recognition has been proposed which consists of two models: 1) A detector, a CNN architecture for detection of gestures; and 2) A classifier, a CNN for classification of the detected gestures. By using dynamic hand gesture recognition to interact with the system, the interactions can be increased with the help of multidimensional use of hand gestures as compared to other input methods. The dynamic hand gesture recognition method focuses to replace the mouse for interaction with the virtual objects. This work centralises the efforts of implementing a method that employs computer vision algorithms and gesture recognition techniques for developing a low-cost interface device for interacting with objects in the virtual environment such as screens using hand gestures.

Traffic Sign Recognition Using Small-Scale Convolutional Neural Network

Traffic sign recognition has been of utmost importance ever since the emergence of the need for autonomous vehicles and driver assistance systems. An effective pre-processing of the data is important in the autonomous driving system. There is no scope to apply complex transformations or highly computational image processing techniques for such real-time purposes. This work presents an approach to recognize traffic signs using small-scale deep convolutional neural networks (CNN) and that can be applied to different applications. The presented solution is implemented using the German Traffic Sign Recognition Benchmark (GTSRB) dataset. This dataset is reliable, vibrant, and has been used for the training of different systems. The proposed system is an Advanced Driver Assistance System (ADAS) based solution to provide effective assistance. The achieved testing, training and validation accuracies are 97.71 %, 99.19%, and 99.61 % respectively.

A Decisive Object Detection using Deep Learning Techniques

Object detection is one of the essential features of computer vision and image processing techniques. In today's world, the computer can replicate or outperform the operation that a human can do. One such thing is object detection, and In the case of it, the machines must be trained in such a way that it can recognize the object equivalent to the human does with maximum accuracy. Several object detection techniques are used to train the machine to detect the objects. Some of the most common object detection techniques are R-CNN, Fast R-CNN, Faster R-CNN) Single Shot MultiBox Detector (SSD), and You Only Look Once(YOLO),. Each of these techniques has a different way of approach and accuracy of detecting the objects in real-time. These techniques are differentiated based on their performances, i.e., speed and accuracy. Some techniques may be very accurate in detecting the objects but may lack in the time taken for detecting the objects, whereas, on the other hand, some techniques may be very fast in figuring out the objects but not with greater accuracy. We have trained an object detection model based on the YOLO technique which gave the best performance out of all other existing techniques, though the accuracy of the model is less, the speed of detection is extremely high. So based on our research we have figured out the best performance object detection techniques and also the most accurate technique. A well-trained object detection model must be very optimistic in terms of their speed and accuracy.

Automatic cephalometric landmarks detection on frontal faces: An approach based on supervised learning techniques

Facial landmarks are employed in many research areas, including facial recognition, craniofacial identification, age and sex estimation being the most important. In forensics, the focus is on the analysis of a particular set of facial landmarks, defined as cephalometric landmarks. Previous studies demonstrated that the descriptive adequacy of these anatomical references for indirect application (photo-anthropometric description) increased the marking precision of these points, contributing to greater reliability of these analyses. Nevertheless, most are performed manually and all are subject to bias on the part of expert examiners. Therefore, the purpose of this work was to develop and validate automatic techniques for detection of cephalometric landmarks from digital images of frontal facial images in forensics. The presented approach uses a combination of computer vision and image processing techniques within supervised learning procedures. The proposed methodology obtains similar precision to a group of human manual cephalometric reference markers and results that are more accurate than other state-of-the-art facial landmark detection frameworks. It achieves a normalized mean distance (in pixels) error of 0.014, similar to the mean inter-expert dispersion (0.009) and clearly better than other automatic approaches that were analyzed during the course of this study (0.026 and 0.101).

Recent Development on Detection Methods for the Diagnosis of Diabetic Retinopathy

Diabetic retinopathy (DR) is a complication of diabetes that exists throughout the world. DR occurs due to a high ratio of glucose in the blood, which causes alterations in the retinal microvasculature. Without preemptive symptoms of DR, it leads to complete vision loss. However, early screening through computer-assisted diagnosis (CAD) tools and proper treatment have the ability to control the prevalence of DR. Manual inspection of morphological changes in retinal anatomic parts are tedious and challenging tasks. Therefore, many CAD systems were developed in the past to assist ophthalmologists for observing inter- and intra-variations. In this paper, a recent review of state-of-the-art CAD systems for diagnosis of DR is presented. We describe all those CAD systems that have been developed by various computational intelligence and image processing techniques. The limitations and future trends of current CAD systems are also described in detail to help researchers. Moreover, potential CAD systems are also compared in terms of statistical parameters to quantitatively evaluate them. The comparison results indicate that there is still a need for accurate development of CAD systems to assist in the clinical diagnosis of diabetic retinopathy.

PHOTOGRAMMETRIC WIREFRAME AND DENSE POINT CLOUD 3D MODELLING OF HISTORICAL STRUCTURES: THE STUDY OF SULTAN SELIM MOSQUE AND YUSUF AGA LIBRARY IN KONYA, TURKEY

Abstract. The photogrammetry enables to getting high accuracy measurement with low-cost and easy application in documentation of historical structures. The object details are signified with lines in cultural heritage documentation by photogrammetry. The combination of all the lines create 3D wireframe model of the measurement object. In addition, patch surfaces of the wireframe are mapped with the texture from the images for more realistic visualization. On the other hand, the progress on computer vision and image processing techniques is allowing automatically perform the photogrammetric process. A large number of points that are called dense point cloud can be measured from coverage area of multi view images. The dense point cloud represents the object shape with small space measured points while the wireframe photogrammetry is representing the object with lines. In this study these two photogrammetric methods were evaluated with respect to visualization, cost, labour and measurement time through 3D modelling of historical structures of Sultan Selim Mosque and Yusuf Aga Library.

Application of CAD systems for the automatic detection of lung nodules

Abstract Lung cancer is a common type of cancer that requires early diagnosis due to its often fatal consequences. Computer image processing techniques may be useful to increase the speed and accuracy of lung cancer detection. In order to process medical images, computerized tomography images usually are incorporated due to their high resolution and low noise level. In this paper, the application of Computer-Aided Detection systems for the diagnosis of lung cancer has been studied - including preprocessing and segmentation methods, and data analysis techniques. The main goal was to investigate the latest technology for the development of computational diagnostic tools, so as to assist with the acquisition, processing, and analysis of the medical imagery data. However, there are aspects that still require further attention, such as to increase the sensitivity, reduce false positives, and optimize the detection of each type of nodule, even for differing size and shape.

Automatic Classification of Brain Tumor and Alzheimer’s Disease in MRI

Computer vision (CV) and image processing techniques aim at the fast development of medical images diagnoses field. As the specialist takes a long time to diagnose one MRI images, CV techniques and machine learning algorithms make the process faster than the manual way, and these techniques save time and effort. In this paper, we developed an intelligent manner for the detection and classification of brain pathologies like tumors, Alzheimer’s disease (AD), or normal brain images. The proposed algorithm encompasses 4 stages: Magnetic Resonance Imaging (MRI) image acquisition, pre-processing, feature extraction, and classification. In this paper, the Bag of Features module has been used for the classification of the MRI of brain with tumor and MRI of brain of Alzheimer’s disease patients from normal brain MRI. Average classification accuracy of 97% is achieved for all three classes.

Quantitative imaging of lipid droplets in single cells.

The combination of next generation sequencing (NGS) and automated liquid handling platforms has led to a revolution in single-cell genomic studies. However, many molecules that are critical to understanding the functional roles of cells in a complex tissue or organs, are not directly encoded in the genome, and therefore cannot be profiled with NGS. Lipids, for example, play a critical role in many metabolic processes but cannot be detected by sequencing. Recent developments in quantitative imaging, particularly coherent Raman scattering (CRS) techniques, have produced a suite of tools for studying lipid content in single cells. This article reviews CRS imaging and computational image processing techniques for non-destructive profiling of dynamic changes in lipid composition and spatial distribution at the single-cell level. As quantitative CRS imaging progresses synergistically with microfluidic and microscopic platforms for single-cell genomic analysis, we anticipate that these techniques will bring researchers closer towards combined lipidomic and genomic analysis.

Pedestrian Recognition Based on Multi-Scale Weighted HOG

Pedestrian recognition receives a great attention in recent years due to its importance in traffic accidents identification. Traffic surveillance systems can provide valuable information for pedestrian recognition using computer vision and image processing techniques. Most techniques exploit simple feature extraction and multi-stage feature matching to train classifiers. In this study, Canny edge information and Histogram of Oriented Gradient (HOG) has been integrated into multi-scale coarse-to-fine feature extraction. Edge distribution provides a variable weight to highlight distinctive gradients in a Multi-Scale Weighted HOG (MS-WHOG) to identify pedestrian. As a result, the pedestrian distinctive features are highlighted and expanded along the edges to improve the recognition process. The proposed technique is also scale and orientation invariant, due to the use of multi-scale and edge information.

Temporal pattern of postmortem color changes in the pupil region of the cornea in rabbits

To explore the temporal pattern of postmortem color changes in the pupil region of the cornea for noninvasive estimation of the postmortem interval (PMI). Two rabbit models of air embolism and drowning were established in a dark room at a temperature of 20 ℃ with a relative humidity of 30%. The corneal images of the rabbits were acquired using a digital camera at two-hour intervals within 72 h after death. The pupil region on the corneal images was segmented using computer image processing technique (MATLAB), and the parameters of 6 image color features (RGBHSV) were extracted. Regression analysis was used to analyze the relationship between these parameters and the PMI, and the effects of different death causes on the changes of the corneal color features were also assessed. Within 72 h after death from different causes, the R, G and B values of the pupil region on the corneal images all tended to increase with the PMI, showing a good fitting with the PMI (P < 0.01). No significant correlation was found between the values of H, S and V and the PMI (P>0.05). The R, G and B values in the pupil region on the corneal images showed consistent variation trends after death from the two causes, and their correlations with PMI were also similar. The measured values of R, G and B in air embolism group were greater than those in the drowning group. The postmortem color changes of the pupil region on corneal images follow an identifiable temporal pattern and can vary across different causes of death. The regression equations established in this study provide references for non-invasive and objective estimation of the PMI.

Linking production and perception of clear speech

Speech communication can adopt different styles as a function of speaking environments and communicative needs. In auditorily or visually challenging contexts, speakers often alter their speech production using a clarified, hyper-articulated speech style with the intention of enhancing speech intelligibility. Such modifications may result in perceptible articulatory and acoustic changes. Questions thus arise as to whether and what clear-speech modifications facilitate perception. This presentation surveys recent research conducted in our labs, investigating clear-speech production and its associated effects on perception. In a series of three-stream studies, this research relates analyses of visible articulatory features using computer image-processing techniques, measurements of acoustic properties, and perceptual patterns of clear-speech segments and suprasegmentals by native and non-native perceivers. Results reveal that clear (relative to plain) speech modulates different and compensatory articulatory-acoustic cues within each sound category to enhance intelligibility. However, our results also show that clear-speech modifications that reduce phonemic category distinctiveness inhibit intelligibility. These findings indicate that clear-speech effects are governed by the collateral principles of within-category cue enhancement and maintenance of category distinctiveness.

Context Semantics for Small Target Detection in Large-Field Images with Two Cascaded Faster R-CNNs

Computer vision and image processing techniques have been widely applied to power transmission line inspection. However, the successful detection of small targets in large scenes is still challenging due to their low resolution and poor feature representation. Existing methods, such as multi-scale image pyramid, multi-scale feature pyramid and multiple heterogeneous feature fusion, etc. can extract more representative features of small objects, but they usually require high computation cost. In this paper, we propose an effective two cascaded Faster R-CNN strategy, which is based on multi-scale features and semantic information between the objects and the background, to address the small target detection in large scenes. Specially, we detect large object candidate proposals that may contain small objects at first and then map them to the original images to detect the small-sized targets on the high resolution regions. Experiments show that our strategy could lead to higher (83.0%) accuracy of small target detection and recognition than the one-stage Faster R-CNN (78.3%) on the dataset of aerial images.

Lymphoma images analysis using morphological and non-morphological descriptors for classification

Mantle cell lymphoma, follicular lymphoma and chronic lymphocytic leukemia are the principle subtypes of the non-Hodgkin lymphomas. The diversity of clinical presentations and the histopathological features have made diagnosis of such lymphomas a complex task for specialists. Computer aided diagnosis systems employ computational vision and image processing techniques, which contribute toward the detection, diagnosis and prognosis of digitised images of histological samples. Studies aimed at improving the understanding of morphological and non-morphological features for classification of lymphoma remain a challenge in this area. This work presents a new approach for the classification of information extracted from morphological and non-morphological features of nuclei of lymphoma images. We extracted data of the RGB model of the image and employed contrast limited adaptive histogram equalisation and 2D order-statistics filter methods to enhance the contrast and remove noise. The regions of interest were identified by the global thresholding method. The flood-fill and watershed techniques were used to remove the small false positive regions. The area, extent, perimeter, convex area, solidity, eccentricity, equivalent diameter, minor axis and major axis measurements were computed for the regions detected in the nuclei. In the feature selection stage, we applied the ANOVA, Ansari-Bradley and Wilcoxon rank sum methods. Finally, we employed the polynomial, support vector machine, random forest and decision tree classifiers in order to evaluate the performance of the proposed approach. The non-morphological features achieved higher AUC and AC values for all cases: the obtained rates were between 95% and 100%. These results are relevant, especially when we consider the difficulties of clinical practice in distinguishing the studied groups. The proposed approach is useful as an automated protocol for the diagnosis of lymphoma histological tissue.

Detection and classification of citrus diseases in agriculture based on optimized weighted segmentation and feature selection

In agriculture, plant diseases are primarily responsible for the reduction in production which causes economic losses. In plants, citrus is used as a major source of nutrients like vitamin C throughout the world. However, ‘Citrus’ diseases badly effect the production and quality of citrus fruits. From last decade, the computer vision and image processing techniques have been widely used for detection and classification of diseases in plants. In this article, we propose a hybrid method for detection and classification of diseases in citrus plants. The proposed method consists of two primary phases; (a) detection of lesion spot on the citrus fruits and leaves; (b) classification of citrus diseases. The citrus lesion spots are extracted by an optimized weighted segmentation method, which is performed on an enhanced input image. Then, color, texture, and geometric features are fused in a codebook. Furthermore, the best features are selected by implementing a hybrid feature selection method, which consists of PCA score, entropy, and skewness-based covariance vector. The selected features are fed to Multi-Class Support Vector Machine (M-SVM) for final citrus disease classification. The proposed technique is tested on Citrus Disease Image Gallery Dataset, Combined dataset (Plant Village and Citrus Images Database of Infested with Scale), and our own collected images database. We used these datasets for detection and classification of citrus diseases namely anthracnose, black spot, canker, scab, greening, and melanose. The proposed technique outperforms the existing methods and achieves 97% classification accuracy on citrus disease image gallery dataset, 89% on combined dataset and 90.4% on our local dataset.

Real-time biscuit tile image segmentation method based on edge detection

In this paper we propose a novel real-time Biscuit Tile Segmentation (BTS) method for images from ceramic tile production line. BTS method is based on signal change detection and contour tracing with a main goal of separating tile pixels from background in images captured on the production line. Usually, human operators are visually inspecting and classifying produced ceramic tiles. Computer vision and image processing techniques can automate visual inspection process if they fulfill real-time requirements. Important step in this process is a real-time tile pixels segmentation. BTS method is implemented for parallel execution on a GPU device to satisfy the real-time constraints of tile production line. BTS method outperforms 2D threshold-based methods, 1D edge detection methods and contour-based methods. Proposed BTS method is in use in the biscuit tile production line.

AUTOMATIC EXTRACTION OF BUILDING BOUNDARIES FROM HIGH RESOLUTION IMAGES WITH ACTIVE CONTOUR SEGMENTATION

Building extraction from remotely sensed images plays an important role in many applications such as updating geographical information system, change detection, urban planning, disaster management and 3D building modeling. Automatic extraction of buildings from aerial images is not an easy task because of background complexity, lighting conditions and vegetation cover that reduces separability or visibility of buildings. As a result, automatic building extraction can be a complex process for computer vision and image processing techniques. In order to overcome this difficulty region-based active contour model was used to automatically detect the boundary of buildings for this study. To extract object boundaries, the model grows or shrinks the initial contour in the image. The main objective of this paper is making active contours algorithm perform without user interaction and to detect automatically initial contours to segment buildings with a software coded in Matlab. This task carried out by morphological operations, band ratio and thresholding methods. In this study, high resolution aerial images with 8 cm ground sampling distance (GSD) were used. Three separate test zones were selected with varying building level of detail on these images. Finally, it was assessed the accuracy of segmented buildings using Correctness, Completeness and Quality metrics by comparing the results images and manually digitized reference image. The proposed approach for building extraction from images was shown to be 98% accurate on buildings with simple geometry and homogeneous roof textures. However accuracy of extracted buildings with heterogeneous roof textures and lighting, and complex geometry is 89%. The results clearly show that automatically calculated initial contour positions work in accordance with the active contour algorithm and easily extraction of the buildings boundaries.

A novel expert system for objective masticatory efficiency assessment.

Most of the tools and diagnosis models of Masticatory Efficiency (ME) are not well documented or severely limited to simple image processing approaches. This study presents a novel expert system for ME assessment based on automatic recognition of mixture patterns of masticated two-coloured chewing gums using a combination of computational intelligence and image processing techniques. The hypotheses tested were that the proposed system could accurately relate specimens to the number of chewing cycles, and that it could identify differences between the mixture patterns of edentulous individuals prior and after complete denture treatment. This study enrolled 80 fully-dentate adults (41 females and 39 males, 25 ± 5 years of age) as the reference population; and 40 edentulous adults (21 females and 19 males, 72 ± 8.9 years of age) for the testing group. The system was calibrated using the features extracted from 400 samples covering 0, 10, 15, and 20 chewing cycles. The calibrated system was used to automatically analyse and classify a set of 160 specimens retrieved from individuals in the testing group in two appointments. The ME was then computed as the predicted number of chewing strokes that a healthy reference individual would need to achieve a similar degree of mixture measured against the real number of cycles applied to the specimen. The trained classifier obtained a Mathews Correlation Coefficient score of 0.97. ME measurements showed almost perfect agreement considering pre- and post-treatment appointments separately (κ ≥ 0.95). Wilcoxon signed-rank test showed that a complete denture treatment for edentulous patients elicited a statistically significant increase in the ME measurements (Z = -2.31, p < 0.01). We conclude that the proposed expert system proved able and reliable to accurately identify patterns in mixture and provided useful ME measurements.

Audio Augmentation for Traffic Signs: A Case Study of Pakistani Traffic Signs

Augmented Reality (AR) extend the appearance of real-world by adding digital information to the scene using computer graphics and image processing techniques. Various approaches have been used to detect, identify and track objects in real environment depending upon the application, shape of the tracking object and environment type. The marker-based tracking technique is the most commonly used method in augmented reality applications in which fiducial markers are put in the real-world for tracking. In this work we proposed a model to detect and identify the traffic signs through marker based technique to improve the usability of marker-based detection in augmented reality applications. We developed an AR application that can detect and recognize the markers designed for Pakistani traffic signs and augment them with voice alert to the driver so that he does prepared for the upcoming hurdle on the road. As identified by literature no work has been performed on augmentation of voice for traffic signs. From experiments the model outperforms than baseline techniques.

Computer Vision and Machine Learning for Viticulture Technology

This paper gives two contributions to the state-of-the-art for viticulture technology research. First, we present a comprehensive review of computer vision, image processing, and machine learning techniques in viticulture. We summarize the latest developments in vision systems and techniques with examples from various representative studies, including, harvest yield estimation, vineyard management and monitoring, grape disease detection, quality evaluation, and grape phenology. We focus on how computer vision and machine learning techniques can be integrated into current vineyard management and vinification processes to achieve industry relevant outcomes. The second component of the paper presents the new GrapeCS-ML database which consists of images of grape varieties at different stages of development together with the corresponding ground truth data (e.g., pH and Brix) obtained from chemical analysis. One of the objectives of this database is to motivate computer vision and machine learning researchers to develop practical solutions for deployment in smart vineyards. We illustrate the usefulness of the database for a color-based berry detection application for white and red cultivars and give baseline comparisons using various machine learning approaches and color spaces. This paper concludes by highlighting future challenges that need to be addressed prior to successful implementation of this technology in the viticulture industry.