Image Processing Stage Research Articles

A Surface Defect Detection Method for Cold-rolled Strip Steel Based on Improved YOLOv8

Cold-rolled strip steel has excellent mechanical properties, and it is widely used in the fields such as automotive manufacturing, construction industry, aircraft manufacturing and electronic product manufacturing. However, many different processing defects occur in the manufacturing process of cold-rolled strip steel, and this article proposes a visual detection method based on YOLOv8 to identify the surface defects on cold-rolled strip steel more accurately. Firstly, several common types of defects in cold-rolled strip steel were analyzed. Then, in order to enrich the dataset for subsequent detection, wavelet transform was used for filtering, denoising, and canny edge detection in the image processing stage for obtaining the clearer input model image with the more prominent graphics. Lastly, the CA attention mechanism was added into the YOLOv8 which is useful to improve the features recognition ability of the model, at the same time, the original standard convolution was replaced by DCS convolution and the loss function was optimized which is useful to get the lightweight models. The improved YOLOv8 model was validated on the NEU-DET dataset, the overall detection performance of the model such as the recall, the accuracy, and the average precision has significantly improved.

A Modified of Fourth-Order Partial Differential Equations Model Based on Isophote Direction to Noise Image Removal

Image denoising is one of the initial stages of image processing. Many models based on the diffusion method have been used to smooth the image. One of the problems, we face in the model based on the diffusion method is its possible loss of edges. The diffusion force is known to be more effective in areas of high frequency. So This paper suggests combining the direction of isophote and fourth-order partial differential equations to reduce the problem of loss edges and preserve important details of the image. The direction of the isophote can regulate the direction of diffusion. Thus, we have a proposed model that can remove the noise in the area while preserving the important edges and details of the image. We have proven the efficiency and superiority of the proposed model by applying it to a set of images and solving it numerically using the finite difference method (FDM).

Integrating deep learning for accurate gastrointestinal cancer classification: a comprehensive analysis of MSI and MSS patterns using histopathology data

AbstractEarly detection of microsatellite instability (MSI) and microsatellite stability (MSS) is crucial in the fight against gastrointestinal (GI) cancer. MSI is a sign of genetic instability often associated with DNA repair mechanism deficiencies, which can cause (GI) cancers. On the other hand, MSS signifies genomic stability in microsatellite regions. Differentiating between these two states is pivotal in clinical decision-making as it provides prognostic and predictive information and treatment strategies. Rapid identification of MSI and MSS enables oncologists to tailor therapies more accurately, potentially saving patients from unnecessary treatments and guiding them toward regimens with the highest likelihood of success. Detecting these microsatellite status markers at an initial stage can improve patient outcomes and quality of life in GI cancer management. Our research paper introduces a cutting-edge method for detecting early GI cancer using deep learning (DL). Our goal is to identify the optimal model for GI cancer detection that surpasses previous works. Our proposed model comprises four stages: data acquisition, image processing, feature extraction, and classification. We use histopathology images from the Cancer Genome Atlas (TCGA) and Kaggle website with some modifications for data acquisition. In the image processing stage, we apply various operations such as color transformation, resizing, normalization, and labeling to prepare the input image for enrollment in our DL models. We present five different DL models, including convolutional neural networks (CNNs), a hybrid of CNNs-simple RNN (recurrent neural network), a hybrid of CNNs with long short-term memory (LSTM) (CNNs-LSTM), a hybrid of CNNs with gated recurrent unit (GRU) (CNNs-GRU), and a hybrid of CNNs-SimpleRNN-LSTM-GRU. Our empirical results demonstrate that CNNs-SimpleRNN-LSTM-GRU outperforms other models in accuracy, specificity, recall, precision, AUC, and F1, achieving an accuracy of 99.90%. Our proposed methodology offers significant improvements in GI cancer detection compared to recent techniques, highlighting the potential of DL-based approaches for histopathology data. We expect our findings to inspire future research in DL-based GI cancer detection.

A deep learning LSTM-based approach for AMD classification using OCT images

Age-related macular degeneration (AMD) is an age-related, persistent, painless eye disease that impairs central vision. The central area (macula) of the retina, located at the back of the eye, sustains damage that is the cause of loss of vision. The early detection of AMD can increase the probability of treatment and prevent vision loss. The AMD can be classified into dry and wet AMD based on the absence of neovascularization. This study introduces a new methodology for the classification of AMD using optical coherence tomography (OCT) retinal images. The proposed methodology is based on three stages. The first stage is the data preparation stage for resizing and normalizing the used images. The second stage is the image processing stage for enhancing the image quality as contrast and resolution these enhancements have been checked by the weighted peak signal-to-noise ratio (WPSNR) methodology. The third stage is the deep feature extraction and classification stage, which consists of two sub-models. The first model is MobileNet V1 which has been used as a deep feature extractor. The second model is LSTM (long short-term memory), fed with deep features to classify the AMD stages. A multi-classification with six separate trials has been employed with the proposed methodology, and compared with other models like DenseNet201 and InceptionV3. The proposed model has been tested on a sample of benchmark data with 4005 grayscale images labeled into three classes. The proposed methodology has achieved an accuracy of 98.85%, a sensitivity of 99.09%, and a specificity of 99.1%. To ensure the effectiveness of the proposed methodology, a comparative analysis has been established with previous approaches in the related field, and the results demonstrated the superiority of the proposed system in AMD multi-classification.

Efficient artificial intelligence approaches for medical image processing in healthcare: comprehensive review, taxonomy, and analysis

In healthcare, medical practitioners employ various imaging techniques such as CT, X-ray, PET, and MRI to diagnose patients, emphasizing the crucial need for early disease detection to enhance survival rates. Medical Image Analysis (MIA) has undergone a transformative shift with the integration of Artificial Intelligence (AI) techniques such as Machine Learning (ML) and Deep Learning (DL), promising advanced diagnostics and improved healthcare outcomes. Despite these advancements, a comprehensive understanding of the efficiency metrics, computational complexities, interpretability, and scalability of AI based approaches in MIA is essential for practical feasibility in real-world healthcare environments. Existing studies exploring AI applications in MIA lack a consolidated review covering the major MIA stages and specifically focused on evaluating the efficiency of AI based approaches. The absence of a structured framework limits decision-making for researchers, practitioners, and policymakers in selecting and implementing optimal AI approaches in healthcare. Furthermore, the lack of standardized evaluation metrics complicates methodology comparison, hindering the development of efficient approaches. This article addresses these challenges through a comprehensive review, taxonomy, and analysis of existing AI-based MIA approaches in healthcare. The taxonomy covers major image processing stages, classifying AI approaches for each stage based on method and further analyzing them based on image origin, objective, method, dataset, and evaluation metrics to reveal their strengths and weaknesses. Additionally, comparative analysis conducted to evaluate the efficiency of AI based MIA approaches over five publically available datasets: ISIC 2018, CVC-Clinic, 2018 DSB, DRIVE, and EM in terms of accuracy, precision, Recall, F-measure, mIoU, and specificity. The popular public datasets and evaluation metrics are briefly described and analyzed. The resulting taxonomy provides a structured framework for understanding the AI landscape in healthcare, facilitating evidence-based decision-making and guiding future research efforts toward the development of efficient and scalable AI approaches to meet current healthcare needs.

Discrete Curvelet Transform Feature Extraction for Mangosteen Fruit Surface Damage Detection

Mangosteen (Garcinia mangostana L) is one of the commodities of Indonesian fruit and is used as an export primadona that became the basis of Indonesia to increase the currency of the country. The quality of the fruit can be seen from the surface, whether there is damage or not. The sorting that the farmers have been doing all this time is still using the conventional way, that is, with the sense of sight. This conventional method seems to be less effective because it takes a lot of energy, takes a long time, and there are different perceptions between farmers. To solve this problem, a method of surface quality extraction of mango fruit will be developed based on image processing. The initial stage of image processing is with the image size equation then the image is converted to grayscale mode, then a discrete curvelet transformation is performed. The next stage is the extraction of mean, energy, entropy, standard deviation, variance, sum, correlation, contrast, and homogeneity. The result of the subsequent feature extraction is used to enter a value at the classification stage. From some of these extractions it will be known which extraction has the highest accuracy value. The method of classification used is Linear Discriminant Analysis (LDA) with the method of K-Fold Cross Validation which in this study is divided into 4-fold cross validation. After testing on 120 images, the highest value of accuracy is with extraction of standard characteristics deviation of 91.7% and variance of 88.4%.

Convolutional neural network and glioma genotyping: A systematic review.

e14038 Background: Gliomas are the most common primary intracranial tumors. Overall, there are 23 different glioma types, and the grading and prognosis of each type are determined by their specific mutations. Several identified mutations are IDH gene status, CDKN2A/B homozygote delegation, PTEN mutation, p53 mutation, TERT promotor mutation, H3K27M mutation, MGMT promotor methylation, trisomy 7/monozomy10, EGFR amplification, ATRX mutation, 1p/19q codeletion, and BRAF mutation. These specific mutations and genomic features can have their distinctive radiologic features seen in different magnetic resonance imaging (MRI) sequences. By using a Convolutional Neural Network (CNN) for feature extraction, the genotyping power of MRI for glioma could be increased. To address this, we systematically reviewed the use of different CNN-based models to detect the genetic context of glioma based on the MRI images. Methods: We conducted a systematic search in PubMed, Embase, Scopus, Web of Science, and Google Scholar. We included original peer-reviewed human studies (retrospective or prospective), focusing on the use of CNN in individuals with glioma. We excluded animal studies, studies with no CNN-based models, reviews, book chapters, conference abstracts, and case reports. JBI checklists were utilized for quality assessment. Results: The initial search resulted in 973 articles. after removing duplicates, 563 studies entered the screening process by title and abstract. 59 studies met our inclusion criteria (56 retrospective and 3 prospective cohort studies) with a total of 26434 glioma patients included (studies ranged from 21 – 2648 individuals). 31 studies used CNN methods only in different stages of image processing (tumor segmentation, feature extraction, image encoder) and 28 studies used CNN as a classifier and mutation predictor. ResNet was the most commonly used CNN-based model (n = 14). U-net (n = 7) and DenseNet models (n = 4) were other commonly used CNN-based models. Included studies targeted IDH status (n=44), 1p/19q codeletion (n=18), MGMT promotor methylation (n=11), ATRX mutation (n=5), TERT promotor mutation (n=5), CDKN2A/B homozygote deletion (n=3), PTEN mutation (n=3), p53 mutation (n=3), H3K27M mutation (n=2), trisomy 7/monozomy10 (n=2), EGFR amplification (n=2), and BRAF mutation (n=1). Highly severe methodological heterogeneity among these studies excluded the possibility of meta-analysis. Conclusions: Current studies have shown that, as a non-invasive method, using CNN-based models on multiple MRI sequences increases the power of early detection and prediction of glioma genetic types before pathological confirmation and has promising results. However, DL performance varies among mutations. Combining multiple DL models with clinical and radiomic features in different and larger databases can further increase the final results.

Digital Image and Video Processing: Algorithms and Applications

Many of the techniques that are used in digital image and video processing were developed in the 1960s at Bell Laboratories. These techniques have applications in a variety of fields, including medical imaging, videophone, character recognition, satellite imagery, and wire-photo standards conversion. Additional applications include enhancement of photographs or vidoes. The early stages of image and video processing were developed with the intention of enhancing the overall quality of the image or video. For the purpose of enhancing the visual effect of humans, it is intended for human beings. When it comes to image and video processing, the input is an image of poor quality, and the output is an image and video of higher quality. Research on algorithms and applications of digital image and video processing is the primary purpose of this study, which aims to investigate these topics extensively. The methodology employed in this study is qualitative research technique. In accordance with the findings of this research, "Image Processing" refers to the process of analyzing images with the objective of determining the significance of objects and identifying them. Image analysts analyze data that has been remotely sensed and attempt to detect, identify, classify, measure, and evaluate the significance of physical and cultural objects, as well as their patterns and spatial relationships. One subcategory of signal processing is known as video processing, and it is distinguished by the fact that the signals that are input and output are video files or video streams. Technology such as television sets, videocassette recorders (VCRs), DVD players, and other devices all make use of video processing algorithms. The processing of images and videos is extremely useful in a variety of contexts.

DEVELOPMENT OF A VISUAL ODOMETRY MODEL BASED ON SENSORS AND VIDEO STREAM ANALYSIS

The article is devoted to the development of a visual odometry model based on sensors of an inertial measuring device and the analysis of a video stream arriving in real time. Modeling is based on the analysis and evaluation of methods for measuring the correct coordinates of a moving object, systems for estimating the movement of an object in three-dimensional space, algorithms at intermediate stages of image processing, principles for selecting special points on the frame and optical flow for selected points.

Preference aggregation method in determining brightness threshold values for object recognition on optical images

Relevance. The need to automate the processing large volumes of geophysical information extracted from images obtained through aerial photography or space remote sensing. The main stage of image processing for object recognition is segmentation. It consists in dividing the image into two homogeneous, clearly distinguishable non-overlapping areas with known boundaries. Threshold segmentation methods are simple but effective and therefore popular. Aim. To describe the experimental studies results of application of the robust method of interval fusion with preference aggregation previously developed by the authors for selecting threshold values when segmenting objects in optical images. Objects. Coastline images taken by Earth remote sensing satellites. Methods. The analyzed image is divided into equal bands. Based on the brightness histogram of each band, its characteristic brightness interval is determined. For the obtained brightness intervals, the fusion result is calculated using the interval fusion with preference aggregation method. The latter is used as a threshold brightness value when forming a segmented image. Results. The results of experimental studies of 100 satellite images of the coastline showed that the proposed method provided correct separation of land and sea regions in 84 images. And the traditional methods of maximum entropy, arithmetic averaging and Otsu showed correct results only in 27, 76 and 73 cases, respectively. Evaluations of such the metrics as Precision, Recall and deviation of the calculated threshold from a known correct value showed the dominance of the proposed method over other tested ones in terms of segmentation quality.

Detection Method for Rice Seedling Planting Conditions Based on Image Processing and an Improved YOLOv8n Model

In response to the need for precision and intelligence in the assessment of transplanting machine operation quality, this study addresses challenges such as low accuracy and efficiency associated with manual observation and random field sampling for the evaluation of rice seedling planting conditions. Therefore, in order to build a seedling insertion condition detection system, this study proposes an approach based on the combination of image processing and deep learning. The image processing stage is primarily applied to seedling absence detection, utilizing the centroid detection method to obtain precise coordinates of missing seedlings with an accuracy of 93.7%. In the target recognition stage, an improved YOLOv8 Nano network model is introduced, leveraging deep learning algorithms to detect qualified and misplaced seedlings. This model incorporates ASPP (atrous spatial pyramid pooling) to enhance the network’s multiscale feature extraction capabilities, integrates SimAM (Simple, Parameter-free Attention Module) to improve the model’s ability to extract detailed seedling features, and introduces AFPN (Asymptotic Feature Pyramid Network) to facilitate direct interaction between non-adjacent hierarchical levels, thereby enhancing feature fusion efficiency. Experimental results demonstrate that the enhanced YOLOv8n model achieves precision (P), recall (R), and mean average precision (mAP) of 95.5%, 92.7%, and 95.2%, respectively. Compared to the original YOLOv8n model, the enhanced model shows improvements of 3.6%, 0.9%, and 1.7% in P, R, and mAP, respectively. This research provides data support for the efficiency and quality of transplanting machine operations, contributing to the further development and application of unmanned field management in subsequent rice seedling cultivation.

Integration of Remote Sensing and Aerogeophysical Data Applied in the Geological Characterization of the Ipueira-Medrado Segment, Andorinhas/BA

The segment of the Ipueira-Medrado mines constitutes two of the main chromium mines operated by Companhia Ferro-ligas da Bahia (Ferbasa). The research area is located in the municipality of Andorinhas, in the north center of the state of Bahia, in the Piemonte Norte Do Itapicuru region. Therefore, the main objective of the present work is the use of Remote Sensing and Aerogeophysics data to elaborate the geological map of the study area and highlight new target areas for mineral research. The Digital Image Processing (PDI) stage was carried out in an environment in the ENVI® 5.3 software for the scenes of the optical Alos and radar images, Palsar. The treatment of magnetic and radiometric data was carried out in the environment of the software Oasis Montaj 9.10 by Geosoft®. In the Geographic Information System (GIS) platform of ArcGis 10.7.1®, multisource data was integrated. This integration supported the screening of information from local lithologies and the identification of four lithological units: Sienito Itiúba, Santa Luz Unit, São Bento Unit and Jacurici Valley Complex. In this last unit, a combined analysis between Map Algebra and Fuzzy logic was applied to all input raster data, in order to mark out a zoning of potential areas for the occurrence of mafic-ultramafic rocks. In the end, a good relationship was noted between the results obtained with models that already exist in the bibliography. In view of this, the application of the methodology developed in this research requires more robustness for the local geological knowledge, as it will also help in the development of new mineral research endeavors.

DAGAN: A GAN Network for Image Denoising of Medical Images Using Deep Learning of Residual Attention Structures

Medical images are susceptible to noise and artifacts, so denoising becomes an essential pre-processing technique for further medical image processing stages. We propose a medical image denoising method based on dual-attention mechanism for generative adversarial networks (GANs). The method is based on a GAN model with fused residual structure and introduces a global skip-layer connection structure to balance the learning ability of the shallow and deep networks. The generative network uses a residual module containing channel and spatial attention for efficient extraction of CT image features. The mean square error loss and perceptual loss are introduced to construct a composite loss function to optimize the model loss function, which helps to improve the image generation effect of the model. Experimental results on the LUNA dataset and “the 2016 Low-Dose CT Grand Challenge” dataset show that DAGAN achieves the best results in root mean square error (RMSE), structural similarity (SSIM) and peak signal-to-noise ratio (PSNR) when compared to the state-of-the-art methods. In particular, PSNR reaches 31.2308 dB and 27.5265 dB, SSIM reaches 0.9115 and 0.7895, while RMSE is 0.0082 and 0.0112, respectively. This indicates that our method performs better than the state-of-the-art methods in the task of CT image denoising.

Compressing the Geospatial Data of Testing Grounds

The paper discusses an algorithm for obtaining the compressed spatial images of testing grounds, which are described by attribute data. The attributive representations of data make it possible to use the hardware and software resources of geographical information systems (GIS) more economically. The objects, that can be presented by a boundary, are considered as the observable ones. The efficiency of the testing ground description is achieved by representing all its areas by boundaries, encoding the boundaries using the Freeman code, and applying the entropy encoding at the last stage of image processing. At the same time, the compressed set of attributive data can be unpacked into the original image with almost no recovery errors or with acceptable accuracy, but not below the limit set by the specification. The theoretical principles of the method are illustrated by the example of processing the segmented image of the contour. The comparative assessment of the efficiency of compression of the combined method under consideration and that of the entropy encoding method is presented.

Evaluation of different median filters for image processing tasks

Impulsive noise is captured at the acquisition stage of digital image processing. It is an important step in the pre-processing process which is used to improve the image quality and make it more informative. There have been several proposed image-denoising algorithms, each with advantages and disadvantages. This work gives a thorough investigation of the median filter and its various variations for the reduction or elimination of impulsive noise in grey scale images. In terms of usefulness and comparative performance, Standard Median Filter, Adaptive Median Filter and Modified Median Filter are contrasted. On a set of pictures, thorough MATLAB-based simulations have been run to assess the performance of the current algorithms. Peak Signal to Noise Ratio was employed to compare performance in this case (PSNR). In terms of relative statistical performance, the Modified Median Filter (MMF) is the best.

De-Noising of Apple Fruit Images using Dilated Convolution Neural Network model with Mixed Activation Function

In the agricultural sector, fruit diseases contribute to substantial economic losses, emphasizing the need for effective disease detection methods. Image processing techniques play a crucial role in minimizing damage and financial losses. Pre-processing, a vital stage in image processing, involves noise elimination and quality enhancement, laying the groundwork for subsequent tasks such as segmentation, classification, and disease detection. This study focuses on, proposing a novel approach for denoising diseased apple fruit images using the Dilated Convolution Neural Network model with Mixed Activation Function (DCNMAF). Performance evaluation, utilizing metrics like PSNR and SSIM, highlights the superiority of the proposed model over existing methods. These advancements underscore the effectiveness of the proposed DCNMAF model, showcasing notable enhancements in both accuracy metrics.

Dimensionality Reduction Hybrid U-Net for Brain Extraction in Magnetic Resonance Imaging.

In various applications, such as disease diagnosis, surgical navigation, human brain atlas analysis, and other neuroimage processing scenarios, brain extraction is typically regarded as the initial stage in MRI image processing. Whole-brain semantic segmentation algorithms, such as U-Net, have demonstrated the ability to achieve relatively satisfactory results even with a limited number of training samples. In order to enhance the precision of brain semantic segmentation, various frameworks have been developed, including 3D U-Net, slice U-Net, and auto-context U-Net. However, the processing methods employed in these models are relatively complex when applied to 3D data models. In this article, we aim to reduce the complexity of the model while maintaining appropriate performance. As an initial step to enhance segmentation accuracy, the preprocessing extraction of full-scale information from magnetic resonance images is performed with a cluster tool. Subsequently, three multi-input hybrid U-Net model frameworks are tested and compared. Finally, we propose utilizing a fusion of two-dimensional segmentation outcomes from different planes to attain improved results. The performance of the proposed framework was tested using publicly accessible benchmark datasets, namely LPBA40, in which we obtained Dice overlap coefficients of 98.05%. Improvement was achieved via our algorithm against several previous studies.

E3ID: An efficient end to end person search model

Person Search has recently emerged as a challenging task that aims to jointly solve Pedestrian Detection and Person Re-identification (Re-ID). However, the existing approaches still stay at the image processing stage. In this paper, we proposed E3ID, an efficient end-to-end person search model for video, to better solve the person search problem in real densely populated areas. To speed up the model, a low-quality image filter is proposed to adaptively adjust the sampling frequency of Yolov5 according to the moving speed of pedestrians in the video with minimal computational effort. And the cropped pedestrian images are stored in the gallery library for the Re-ID step. Then, we exploit color feature enhancer to mine color features in specific regions that contribute significantly to the Re-ID process. Compared to other state-of-the-art methods, our proposed E3ID can reach 77.94 % on the evaluation index rank 1, which creates a feasibility of using this model in other complex real-world scenarios.

Image Reconstruction Using an Array of Photoresistors and Kalman Filter

On this paper we present a mechano-electrical system capable of scanning and acquiring images that are enhanced using Kalman filter. As photosensitive element we chose photoresistors that restrict the resolution of our system to their diameter. The whole system consists of three main parts: the scanning device, the acquiring circuit and finally an image processing stage. To process the image, we designed a Kalman Filter, that allows us to save two images, the one acquired directly from the mechano-electrical circuit and the one enhanced by the filter. Our system proposes a simple, small, low cost scanning solution that can be used as a camera for images that have little curves on their morphology.

Light-dependent changes in the outer plexiform layer of the mouse retina.

The ability of the visual system to relay meaningful information over a wide range of lighting conditions is critical to functional vision, and relies on mechanisms of adaptation within the retina that adjust sensitivity and gain as ambient light changes. Photoreceptor synapses represent the first stage of image processing in the visual system, thus activity-driven changes at this site are a potentially powerful, yet under-studied means of adaptation. To gain insight into these mechanisms, the abundance and distribution of key synaptic proteins involved in photoreceptor to ON-bipolar cell transmission were compared between light-adapted mice and mice subjected to prolonged dark exposure (72 hours), by immunofluorescence confocal microscopy and immunoblotting. We also tested the effects on protein abundance and distribution of 0.5-4 hours of light exposure following prolonged darkness. Proteins examined included the synaptic ribbon protein, ribeye, and components of the ON-bipolar cell signal transduction pathway (mGluR6, TRPM1, RGS11, GPR179, Goα). The results indicate a reduction in immunoreactivity for ribeye, TRPM1, mGluR6, and RGS11 following prolonged dark exposure compared to the light-adapted state, but a rapid restoration of the light-adapted pattern upon light exposure. Electron microscopy revealed similar ultrastructure of light-adapted and dark-adapted photoreceptor terminals, with the exception of electron dense vesicles in dark-adapted but not light-adapted ON-bipolar cell dendrites. To assess synaptic transmission from photoreceptors to ON-bipolar cells, we recorded electroretinograms after different dark exposure times (2, 16, 24, 48, 72 hours) and measured the b-wave to a-wave ratios. Consistent with the reduction in synaptic proteins, the b/a ratios were smaller following prolonged dark exposure (48-72 hours) compared to 16 hours dark exposure (13-21%, depending on flash intensity). Overall, the results provide evidence of light/dark-dependent plasticity in photoreceptor synapses at the biochemical, morphological, and physiological levels.