Image Processing Stage Research Articles

A Medically Assisted Model for Precise Segmentation of Osteosarcoma Nuclei on Pathological Images.

Osteosarcoma is the most common malignant bone tumor with a high degree of malignancy and misdiagnosis rates. Pathological images are crucial for its diagnosis. However, underdeveloped regions currently lack sufficient high-level pathologists, leading to uncertain diagnostic accuracy and efficiency. Existing research on pathological image segmentation often neglects the differences in staining styles and lack of data, without considering medical backgrounds. To alleviate the difficulty in diagnosing osteosarcoma in underdeveloped areas, an intelligent assisted diagnosis and treatment scheme for osteosarcoma pathological images, ENMViT, is proposed. ENMViT utilizes KIN to achieve normalization of mismatched images with limited GPU resources and uses traditional data enhancement methods, such as cleaning, cropping, mosaic, Laplacian sharpening, and other techniques to alleviate the issue of insufficient data. A multi-path semantic segmentation network combining Transformer and CNN is used to segment images, and the degree of edge offset in the spatial domain is introduced into the loss function. Finally, noise is filtered according to the size of the connecting domain. This paper experimented on more than 2000 osteosarcoma pathological images from Central South University. The experimental results demonstrate that this scheme performs well in each stage of the osteosarcoma pathological image processing, and the segmentation results' IoU index is 9.4% higher than the comparative models, demonstrating its significant value in the medical industry.

From Imaging Data to Cranioplasty Implant Designs

The cranioplasty procedure is starting from removal the skull bone defects and replacing them with any biocompatible material, such as polymer, ceramic, or titanium alloy. The complication of the surgery as well as the high cost from several material selection required a simulation. Besides that, the case of cranial defects sometimes required a customized design. The presence of three-dimensional (3D) printing technology would be a promising tool to improve the success rate. Prior to 3D printing, the model needs to be corrected from the initial patient’s imaging data to the intended implant design. However, previous related literatures were almost not informing the specific image processing steps to gain the models, while not all operators could understand this sophisticated technique. The study aims to design an implant bone for cranioplasty purpose. The data were processed through the very clear step-by-step image processing stages, three-dimensional (3D) printing, and its evaluation through biomechanical simulation. Quantitatively, the designed cranioplasty implant could deal with the load in the actual application. Qualitatively, the prototypes have matched if applied to the host of cranium bone. In conclusion, although image processing and refinements are the most complicated process, the whole explanation indicate that the provided precise methodology could be a major reference to the similar procedure.

Development of convolutional neural network algorithm on ships detection in Natuna Islands-Indonesia using land look satellite imagery

Indonesia's vast waters require much time and energy to monitor illegal fishing. Utilizing remote sensing technology is the right solution for monitoring, controlling, and surveying (MCS) systems. The phenomenon of particular concern to the ship detection process using a remote sensing system is the form of images with various positions due to the ship's rectangular and narrow dimensions and the high viewing angle from space. This condition differs if the shooting is carried out close to objects on the earth's surface. Considering the situation of recording data from a very far position, refining the convolutional neural network (CNN) method for processing remote sensing image data continues to be developed, primarily related to detecting ships. The latest development of the CNN algorithm shows its ability to see vessels by taking into account the orientation factor of objects at sea level with almost perfect accuracy. The problem is that the system needs to be able to detect in detail the movements and types of ships recorded by remote sensing satellite sensors. Therefore, an automatic ship detection instrument is needed based on Land look USGS base map high-resolution image data to maintain maritime sovereignty in Indonesia. Image data were acquired from the waters of the Natuna islands and taken at a scale of 1: 2000. The data obtained from the acquisition are five high-resolution image data 1680 × 840 pixels, in which there are 32 ship image data and 3100 non-ship image data 80 × 80 pixels. The ship image data is then processed by color and geometry augmentation to produce 1440 new data. The image processing stage afterward is reading and splitting the dataset, creating a CNN, and training and testing the CNN model. Based on the results of the ship and non-ship objects detection with the CNN model, the accuracy results are 99.78% and 99.24%, respectively.

DEVELOPMENT OF A NEW ONE-EYE IMPLANT BY 3D BIOPRINTING TECHNIQUE

The paper analyzes the possibility of implementing a new eye implant using the 3D bioprinting technique and the experimental model underlying it. The paper includes the stages of image processing using Digital Imaging and Communications in Medicine (DICOM) and the processing of the eye implant, results, and exploitation. In the exploitation part of the eye implant, the aims are optimizing the image processing stages and refining the precision of the calibration model. In the manufacturing part, the new eye implant is fabricated with the help of machines. These machines are suitable for manufacturing structures with their porosity controlled by additive manufacturing techniques. The feasibility of using 3D printing techniques using biocompatible materials in structures with predetermined porosity was demonstrated to manufacture a customized eye implant using medical imaging Computed Tomography (CT) results. The work was performed within the OrbImplant Project.

Quadrotor UAV Dynamic Visual Servoing Based on Differential Flatness Theory

In this paper, we propose 2D dynamic visual servoing (Dynamic IBVS), where a quadrotor UAV tries to track a moving target using a single facing-down perspective camera. As an application, we propose the tracking of a car-type vehicle. In this case, data related to the altitude and the lateral angles have no importance for the visual system. Indeed, to perform the tracking, we only need to know the longitudinal displacements (along the x and y axes) and the orientation along the z-axis. However, those data are necessary for the quadrotor’s guidance problem. Thanks to the concept of differential flatness, we demonstrate that if we manage to extract the displacements according to the three axes and the orientation according to the yaw angle (the vertical axis) of the quadrotor, we can control all the other variables of the system. For this, we consider a camera equipped with a vertical stabilizer that keeps it in a vertical position during its movement (a gimbaled camera). Other specialized sensors measure information regarding altitude and lateral angles. In the case of classic 2D visual servoing, the elaboration of the kinematic torsor of the quadrotor in no way guarantees the physical realization of instructions, given that the quadrotor is an under-actuated system. Indeed, the setpoint has a dimension equal to six, while the quadrotor is controlled only by four inputs. In addition, the dynamics of a quadrotor are generally very fast, which requires a high-frequency control law. Furthermore, the complexity of the image processing stage can cause delays in motion control, which can lead to target loss. A new dynamic 2D visual servoing method (Dynamic IBVS) is proposed. This method makes it possible to generate in real time the necessary movements for the quadrotor in order to carry out the tracking of the target (vehicle) using a single point of this target as visual information. This point can represent the center of gravity of the target or any other part of it. A control by flatness has been proposed, which guarantees the controllability of the system and ensures the asymptotic convergence of the generated trajectory in the image plane. Numerical simulations are presented to show the effectiveness of the proposed control strategy.

Random Forests Algorithm for Two Levels of Coral Reef Ecosystem Mapping Using Planetscope Image in Malalayang Beach, Manado

The coral reef ecosystem has a significant physical and biological function and is also one of the coastal ecosystem components apart from the seagrass and mangrove ecosystem. Besides their ecological function, the coral reef also has an economic function. The condition of the coral reef ecosystem in Malalayang Beach has been changing for years. The utilization of remote sensing images can monitor current conditions. This research aims to map the coral reef ecosystem mapping in Malalayang Beach, Manado and conduct a test for the accuracy of coral reef ecosystem mapping using field survey data as a classification and validation sample. PlanetScope multispectral image has four channels to detect underwater objects: red, green, blue and near infrared. PlanetScope level 3B image for the research has a surface reflectance value for its pixel. The image processing stages of this research consist of sunglint correction, water column correction, and then continue to classify the coral reef ecosystem using random forests algorithm. Classification and accuracy training sample data were obtained using the photo transect technique. The sunglint correction regression equation is between 0.27 – 0.38. The coefficient of attenuation ratio in B1 is 0.927797938, B2 is 0.168841585, and B3 is 0.29033029. This value then becomes the input for the Lyzenga formula. The classification accuracy for level one using random forests is 72,54%, and the accuracy for level two mapping is 37,61%.Keywords: Coral Reef Ecosystem, Planetscope, Random Forests

A low cost IoT-based Arabic license plate recognition model for smart parking systems

License Plate Recognition is one of the significant enablers that can be utilized in wide range of applications in ITS and smart cities. The proposed design relies on three image processing stages to achieve license plate identification with high accuracy which are pre-processing, segmentation, and character recognition. The canny edge detection method with various thresholds, contour detection, and masking techniques are used to locate the car edges and license plate. In the experiment presented in this paper, 200 images were used to identify Egyptian car plates. The model successfully identified Arabic license plates with 93% accuracy. A prototype is implemented using ESP32 Cameras and Raspberry-Pi to test the system's performance. Moreover, a database and a website are hosted on the RPi to allow users to search for their car location in the parking lot using the car's full or partial license plate which was saved in database upon detection.

Моделирование процесса распознавания символов в нормативных документах организации

Improving the quality of classification of different documents is a purpose of modeling the optical character recognition. Non-digital documents, such as scanned or photographed documents, are difficult to classify correctly in electronic document management systems. A decision was made to simulate the process of optical character recognition in the regulatory documents of the organization. There have been considered various methods of modeling the process. The structure of departments for the electronic document management system is given. Methods of implementing optical character recognition (OCR) are considered. The stages of the OCR system development are revealed: image processing, segmentation, recognition. The methods of image processing are analyzed. The main processes associated with image processing are disclosed: alignment, blurring, binarization, finding contours, removing extra lines. Comparison of image blur methods is made. Two stages of image binarization are defined: conversion of a color image into a gray image, binarization of a gray image. The Kenny operator is proposed as a second stage of binarization, which is used to detect the boundaries of the image. The last stage of image processing is the process of removing extra lines. Algorithms for dividing text areas into segments are considered. 3 stages of segmentation are identified: string segmentation, word segmentation, character segmentation. A segmentation algorithm is defined based on calculating the average brightness of image pixels to search for different intervals: line spacing, word spacing, character spacing. Available popular online OCR services as well as some popular desktop programs are considered. A connection has been found between an artificial neural network and optical object recognition. To implement the recognition stage, it is proposed to use an artificial neural network.

Transfer Learning for Image-Based Malware Detection for IoT

The tremendous growth in online activity and the Internet of Things (IoT) led to an increase in cyberattacks. Malware infiltrated at least one device in almost every household. Various malware detection methods that use shallow or deep IoT techniques were discovered in recent years. Deep learning models with a visualization method are the most commonly and popularly used strategy in most works. This method has the benefit of automatically extracting features, requiring less technical expertise, and using fewer resources during data processing. Training deep learning models that generalize effectively without overfitting is not feasible or appropriate with large datasets and complex architectures. In this paper, a novel ensemble model, Stacked Ensemble-autoencoder, GRU, and MLP or SE-AGM, composed of three light-weight neural network models-autoencoder, GRU, and MLP-that is trained on the 25 essential and encoded extracted features of the benchmark MalImg dataset for classification was proposed. The GRU model was tested for its suitability in malware detection due to its lesser usage in this domain. The proposed model used a concise set of malware features for training and classifying the malware classes, which reduced the time and resource consumption in comparison to other existing models. The novelty lies in the stacked ensemble method where the output of one intermediate model works as input for the next model, thereby refining the features as compared to the general notion of an ensemble approach. Inspiration was drawn from earlier image-based malware detection works and transfer learning ideas. To extract features from the MalImg dataset, a CNN-based transfer learning model that was trained from scratch on domain data was used. Data augmentation was an important step in the image processing stage to investigate its effect on classifying grayscale malware images in the MalImg dataset. SE-AGM outperformed existing approaches on the benchmark MalImg dataset with an average accuracy of 99.43%, demonstrating that our method was on par with or even surpassed them.

Applications of imaging systems for the assessment of quality characteristics of bread and other baked goods: A review.

One of the most widely researched topics in the food industry is bread quality analysis. Different techniques have been developed to assess the quality characteristics of bakery products. However, in the last few decades, the advancement in sensor and computational technologies has increased the use of computer vision to analyze food quality (e.g., bakery products). Despite a large number of publications on the application of imaging methods in the bakery industry, comprehensive reviews detailing the use of conventional analytical techniques and imaging methods for the quality analysis of baked goods are limited. Therefore, this review aims to critically analyze the conventional methods and explore the potential of imaging techniques for the quality assessment of baked products. This review provides an in-depth assessment of the different conventional techniques used for the quality analysis of baked goods which include methods to record the physical characteristics of bread and analyze its quality, sensory-based methods, nutritional-based methods, and the use of dough rheological data for end-product quality prediction. Furthermore, an overview of the image processing stages is presented herein. We also discuss, comprehensively, the applications of imaging techniques for assessing the quality of bread and other baked goods. These applications include studying and predicting baked goods' quality characteristics (color, texture, size, and shape) and classifying them based on these features. The limitations of both conventional techniques (e.g., destructive, laborious, error-prone, and expensive) and imaging methods (e.g., illumination, humidity, and noise) and the future direction of the use of imaging methods for quality analysis of bakery products are discussed.

An Integrated Framework with Deep Learning for Segmentation and Classification of Cancer Disease

This paper addresses radiologists’ specific diagnosis of cancer disease effectively using integrated framework of deep learning model. Although several existing diagnosis systems have been adopted by a physician, in few cases, it is not so practical to see the infected area from images in the normal eye. Thus, a fully integrated diagnosis framework for disease detection is proposed to find out the infected area from image using deep learning approaches in this paper. In this proposed framework, various components are designed through deep learning approaches such as detection, segmentation, classification etc. based on mass region. The classification technique is used to classify the disease as either benign or malignant. The vital part of this framework is developed by using a full resolution convolutional network (FrCN) that supports different stages of image processing, especially breast cancer disease. Different experimental evaluation is taken to perform on the accuracy, cross-validation tests, and the comparative testing. Since we have taken 4-fold evaluation, the FrCN performs with an average 98.7% Dice index, 97.8% TS/CSI coefficient, 99.1% overall accuracy, and 98.15% MCC. Our experiments demonstrated that the proposed diagnosis system performs on the deep learning approaches at each segmentation stage and classification with good results.

Correction of rotational blur in images of stars observed by an astroinertial attitude sensor against the background of the daytime sky

A rotational blur correction algorithm is considered as the initial stage of image processing in the problem of attitude measurement using a star tracker. To implement this algorithm, the star tracker must be equipped with a three-axis gyroscope. The algorithm does not guarantee the detection of an image of a star against the background of the daytime sky in one frame but facilitates conditions for subsequent image stacking. The correction aims to localize energy maxima of the blurred star images in pixels with predetermined characteristics. The correction highlights these pixels against the background and improves the signal-to-noise ratio, though deteriorating the artistic quality of the whole digital image. The key characteristic of the pixel of maximum localization is that it is where the geometric image of the star is found at the start of the exposure of the frame under correction. The correction is performed in the form of frame processing with a digital finite-impulse-response (FIR) filter. The impulse response of the filter is inhomogeneous and represents a core of rotational blur, synthesized in each pixel of the corrected frame. Algorithms for calculating levels of the signal, background, and noise in the image of a star observed against the background of the daytime sky with a rotating camera are described. Dependences of the signal-to-noise ratios in various pixels of a blurred image on the exposure time and on the angular velocity of the camera rotation are analyzed. The signal-to-noise ratios in the star image before and after the blur correction are calculated. The simulation results are illustrated by the example of an image of a bright star, clearly showing specific features of the proposed rotational blur correction algorithm.

Image processing and machine learning‐based bone fracture detection and classification using X‐ray images

AbstractOne of the most important problems in orthopedics is undiagnosed or misdiagnosed bone fractures. This can lead to patients receiving an incorrect diagnosis or treatment, which can result in a longer treatment period. In this study, fracture detection and classification are performed using various machine learning techniques using of a dataset containing various bones (normal and fractured). Firstly, the X‐ray images obtained are subjected to image preprocessing stages and are prepared for the feature extraction stage. Then, in addition to the Canny and Sobel edge detection methods used in the image processing stage, feature extraction of X‐ray images is performed with the help of Houhg line detection and Harris corner detector. The data obtained by performing feature extraction is given to 12 different machine learning classifiers and the results are presented. Setting hyperparameters for classifiers is done by the grid search method, and the study is tested for 10‐fold cross‐validation. Classifier results are presented comparatively as accuracy, training time, and testing time, and linear discriminant analysis (LDA) reaches the highest accuracy rate with 88.67% and 0.89 AUC. The proposed computer‐aided diagnosis system (CAD) will reduce the burden on physicians by identifying fractures with high accuracy.

Multi-verse Optimizer with Rosenbrock and Diffusion Mechanisms for Multilevel Threshold Image Segmentation from COVID-19 Chest X-Ray Images.

Coronavirus Disease 2019 (COVID-19) is the most severe epidemic that is prevalent all over the world. How quickly and accurately identifying COVID-19 is of great significance to controlling the spread speed of the epidemic. Moreover, it is essential to accurately and rapidly identify COVID-19 lesions by analyzing Chest X-ray images. As we all know, image segmentation is a critical stage in image processing and analysis. To achieve better image segmentation results, this paper proposes to improve the multi-verse optimizer algorithm using the Rosenbrock method and diffusion mechanism named RDMVO. Then utilizes RDMVO to calculate the maximum Kapur's entropy for multilevel threshold image segmentation. This image segmentation scheme is called RDMVO-MIS. We ran two sets of experiments to test the performance of RDMVO and RDMVO-MIS. First, RDMVO was compared with other excellent peers on IEEE CEC2017 to test the performance of RDMVO on benchmark functions. Second, the image segmentation experiment was carried out using RDMVO-MIS, and some meta-heuristic algorithms were selected as comparisons. The test image dataset includes Berkeley images and COVID-19 Chest X-ray images. The experimental results verify that RDMVO is highly competitive in benchmark functions and image segmentation experiments compared with other meta-heuristic algorithms.

Delimitation of water areas using remote sensing in Brazil’s semiarid region

Remote sensing techniques are of fundamental importance to investigate the changes occurred in the terrestrial mosaic over the years and contribute to the decision-making by increasing efficient environmental and water management. This article aimed to detect, demarcate and quantify the hydric area of Poço da Cruz reservoir, located in Ibimirim, Pernambuco, semiarid region of Brazil, with modeling based on Landsat 8/OLI satellite multispectral images from 2015 to 2020, and to relate it with data from the Climate Hazards Group InfraRed Precipitation with Station (CHIRPS) satellites average rainfall. For this purpose, the Modified Normalized Difference Water Index (MNDWI) was modeled, being produced georeferenced theme maps and extracted only the pixels represented by positive spectral values, which represent water targets. The open-access software Quantum Geographic Information System (QGIS, version 2.18.16) was used for all stages of digital image processing and connection with complementary databases on the theme maps elaboration. In the results, changes in the spatial distribution of Poço da Cruz were evidenced and analyzed using precipitation data from the CHIRPS product, allowing a better understanding of the rainfall behavior in the region and its influence. The MNDWI was lined with the CHIRPS product, in which the spatial correlation between the rainy event and the water area’s delimitation is documented, especially in October 2017 (minimum values) and October 2020 (maximum values).

Використання комп’ютерного зору для візуальної перевірки якості груш

A brief description of the basic stages of image processing is given to pay attention to the segmentation stage as a possible way to improve efficiency in decision-making. The main characteristics of the presented model are visual signs, such as color, shape, the presence of a stem, and others. Due to the different approaches in image processing, a high level of truthfulness is achieved, which is expressed in the percentage ratio of the accuracy of decision-making and varies in the range from 90 to 96%. Therefore, the results obtained in this work make it possible to automate the process of visual inspection with the prospect of increasing the speed and quality of product sales for the consumer.

Моделирование алгоритма повышения разрешающей способности гиперспектральной системы

The article deals with the issues of computer simulation of algorithms for processing spectral images, namely: the procedure for increasing their clarity. For this purpose, a spectral image obtained from a hyperspectrometer is used, as well as additional information from a high-resolution color video camera. The hyperspectrometer and the color video camera form a hyperspectral system in which various processing procedures can be implemented. The article considers a relatively simple processing algorithm with use of E. Land's retinex theory. A block diagram of the image processing stages is presented, as well as the results of computer modeling using real spectral images. Comparative quantitative characteristics of reference and processed images are analyzed. It is shown by modeling that with insignificant spatial distortions, the number of singular points (gradients) of the image, and, consequently, the clarity of the original spectral image increase by several times.

Template Matching Algorithm Implementation For Introduction To Indonesian Traditional House

Digital image processing is an important way in computer vision to determine the shape of digital image objects. Many image processing applications have been produced, one of which is pattern recognition. Pattern recognition can be interpreted as the process of classifying a number of objects into several categories based on similarities and similarities in characteristics. In this study, the pattern recognition of traditional houses in Indonesia will be carried out. In the pattern recognition process, several stages of image processing (image pre-processing) are needed which aims to make the pattern obtained can be recognized accurately. For the introduction of this traditional house, several stages will be carried out before it can finally be recognized by the system. First, the image will be converted into a gray image, after the gray image is obtained, it will be converted into a binary image using thresholding and then followed by normalizing the image size (resize). After that, the resized image will be recognized using the Template Matching method. The results of this study, using 20 examples of traditional house patterns categorized into 10 types of traditional houses in Indonesia, the system can recognize 16 traditional house patterns correctly with an accuracy percentage of 80%. But this system also still has imperfections. Imperfections occur due to several factors such as taking the angle of the object in the test image that does not match the angle of the object in the target image, and also the presence of additional objects that cover the shape of the traditional house that will be recognized; such as: trees, cars, etc., as well as poor test image conditions.

A high-accuracy circular hole measurement method based on multi-camera system

Industrial parts often contain a large number of circular holes that are often used as assembly datums. Measuring circular holes with high accuracy is very challenging due to the low signal-to-noise ratio of the edges on the image. This problem is further complicated when chamfers and nonuniform reflections are present. In this paper, a high-accuracy method for measuring circular holes is proposed to realize on-line inspection under a multi-camera system. In the image processing stage, an ellipse detector is designed based on a high-confidence edge detection algorithm and state-of-the-art ellipse fitting technology to significantly eliminate the ellipse fitting bias caused by blurred edges. In the ellipse matching and validation stage, the ambiguous objects are reconstructed first using ambiguous ellipse pairs matched by the epipolar geometry. Then, the similarity between the reprojected ellipse and the detected ellipse is assessed by calculating the Euclidean distance to significantly eliminate mismatched candidates and select the best matching ones. Subsequently, the discrete 3-D points are calculated by spatial intersection after obtaining the homologous points on the corresponding ellipses. Furthermore, a robust spatial circle fitting method is developed to adapt to discrete 3-D points that are not strictly located on a plane. Finally, extensive experiments on both real-world datasets and simulation datasets demonstrate that the proposed algorithm is beneficial for improving the measurement accuracy compared with the state-of-the-art algorithms and demonstrate great potential for industrial applications.

Exploit computer vision inpainting approach to boost deep learning models

In today’s world, the amount of available information grows exponentially every day. Most of this data is visual data. Correspondingly, the demand for the algorithm of image rent is growing. Traditionally, the first approaches to computer vision problems were classical algorithms without the use of machine learning. Such approaches are limited by many factors. First of all, the conditions imposed on the input images are applied – the shooting angle, lighting, position of objects on the scene, etc. Other classical algorithms cannot meet the needs of modern computer vision problems. Neural network approaches and deep learning models have largely replaced classical programming algorithms. The greatest advantage of deep neural networks in computer vision tasks is not only the possibility of automatically building data processing algorithms that cannot be built in any other way, but also the comprehensiveness of such an approach – actual deep neural networks provide all stages of image processing from start to finish. But. This approach is not always optimal. Training models require a large amount of annotated data to avoid the effect of overfitting such models. In many settings, the conditions have a significant degree of variability, but are limited. In such cases, the combination of both approaches of computer vision is fruitful – pre-processing of the image is performed by classical algorithms, and prediction (classification, object search, etc.) is performed by a neural network. This article noted an example of the use of damaged images in the classification of tasks (in the extreme cases, the percentage of damage reached 60 % of the image area). We have shown in practice that the use of classic approaches for restoration of damaged areas of the image (inpainting) made it possible to increase the final accuracy of the model by up to 10 % compared to the base model trained under identical conditions on the original data.