A revolutionary imaging device termed wireless capsule endoscopy (WCE) provides a painless, noninvasive vision of the complete gastrointestinal (GI) tract. Given the poor intensities and vague differences between bleeding and normal regions, successful bleeding region detection in WCE recordings is essential for the premature diagnosis and restoration of intestinal cancer. Conversely, in this research, we propose a flexible segmentation map construction technique centered on U-shaped architecture and supervised by the suggested Separa-ble Contextual Feature-Guided (SCFG) module to identify GI bleeding. The SCFG module arrives at every encoder-decoder block for guiding the model in accurately recognizing various bleeding zones. The primary drawback is that the module can be improved by better characterizing the bleeding zone due to the usage of numerous alternative convolutional forms. Additionally, a bottleneck module is proposed, serving as a feature transition module. The results of the studies reveal that our suggested system outperforms state-of-the-art segmentation techniques by a margin of 0.47 % in the dice coefficient. The dice score's calculated segmentation rate is 91.14 %, while the intersection over union (IoU) segmentation efficiency is 83.72 %.

The human gastrointestinal (GI) tract is sus-ceptible to a myriad of diseases that can profoundly impact health. Therefore, timely detection and intervention are critical in halting the progression of these diseases and preventing their potential transformation into cancer. Regular screenings, diagnostic tests, and early symptom recognition play vital roles in ensuring early intervention and better patient outcomes in managing GI-related conditions. In recent years, scientists have increasingly turned to advanced technologies, particularly deep learning algorithms, to revolutionize the detection and segmentation of colorectal anomalies. Leveraging the power of artificial intelligence, researchers are exploring sophisticated deep learning models capable of analyzing vast amounts of endoscopic imagery with remarkable precision and efficiency. In our proposed paper, we introduce an approach to the automated segmentation of colorectal anomalies, through the development of an end-to-end architecture named VGGI6U-Net. The architecture enhances the capability of the framework for precise segmentation By leveraging the features of VGG 16 and integrating an attention mechanism into the U-Net framework. The model exhibits promising performance in accurately identifying and delineating polyps and bleeding regions within images. The incorporation of the attention mechanism enables the network to focus on salient features, thereby further improving segmentation accuracy and reducing false positives.

Globally, colorectal cancer is the primary cause of cancer-related death. Colonoscopy is currently one of the most common ways to identify precancerous gastrointestinal disorders. Thus, early and precise polyp segmentation is of therapeutic importance in reducing the risk of developing cancer. The manual examination is a tedious and time-consuming procedure for physicians. Many computer algorithms have been created by scientists to automatically identify problems from colonoscopic images. In order to further increase polyp segmentation performance, we provide in this study an attDenseU-Net design that concurrently includes the attention mechanism and U-Net. AttDenseU-Net reduces the amount of background in an input image while emphasizing key components by inserting a sequence of attention units in between relevant downs amp ling and upsampling operations. Integrating DenseNet blocks into U-Net architecture helps to more effective and efficient feature learning, and improved gradient flow. A publicly available dataset called K vasir-SEG was utilized in this study to evaluate and confirm the proposed approach. Our model's accuracy rate is 86.31 %, its Dice coefficient is 91.48%, and its Jaccard index is 84.30 %. The experiment findings show that the proposed AttDenseU-Net outperforms its baselines and provide s performance on par with existing polyp segmentation methods.

Colorectal cancer is the main reason for mortality from cancer globally. One of the most popular methods for spotting precancerous gut illnesses right now is a colonoscopy. As a result, early and accurate polyp segmentation has great therapeutic significance in lowering the likelihood that cancer may develop. For doctors, the manual examination is a laborious and time-consuming process. In order to automatically separate the abnormalities from endoscopic pictures, numerous computer algorithms have been developed by scientists. In this research, we introduce an attention U-Net architecture that simultaneously incorporates the attention mechanism and U-Net for further performance improvement of polyp segmentation. AttResU-Net inserts a series of attention units between related downsampling and upsampling processes in order to minimize outside regions in an input image while accentuating important elements. This paper used CVC-ClinicDB, a publicly accessible dataset to assess and validate the suggested strategy. Our model has a 97.46% accuracy rate, a 90.85% Dice coefficient, and a 83.26% Jaccard index. The results of the trial demonstrate that the suggested AttentionU-Net surpasses its baselines and offers performance comparable to current polyp segmentation techniques.

Vision Transformer (ViT) and its variants have gained significant prominence in computer vision due to their exceptional performance across various tasks. However, ViTs are data-hungry models that require vast amounts of data for training. Given the scarcity of medical data, this paper presents a fine-tuned vision transformer specifically designed for small-size datasets. We fine-tuned the original model using convolutional neural networks (CNNs) to extract both high and low-level features from wireless capsule endoscopy (WCE) images. In this work, we integrate a CNN module into the original ViT to extract features from WCE patches, which are then fused with the original ViT features. The classification is accomplished using a multilayer perceptron (MLP) to categorize images into normal and abnormal categories on the Kvasir Capsule Endoscopy dataset, as well as bleeding or non-bleeding categories on the Red Lesion Endoscopy dataset. The experimental findings substantiate the efficacy of the suggested approach, yielding favorable outcomes in comparison to other state-of-the-art methods.

Wireless capsule endoscopy (WCE) is a technology that captures images of the digestive tract with a pill-sized camera. Capsule endoscopies are used to exclude or diagnose disorders such as bleeding, early symptoms of gastrointestinal cancer, abdominal pain, Crohn's disease, Celiac disease, polyps, and ulcers. However, the main cause for a capsule endoscopy is to scout for small intestine haemorrhage. Because of the technological limits, the images are low quality and feature multiple orientations due to the capsule's free mobility. In this study, we propose a technique for detecting bleeding in WCE images. We deploy a deep neural network that uses the Inception-ResNet-V2 model for its high level, combined with a low-level model that is a convolutional neural network (CNN), to attain better classification performance. The proposed methods' average accuracy is 98.5 %, with sensitivity, specificity, and precision of 98.5 %, 99 %, and 98.5 %, respectively. It clearly shows that our method outperforms state-of-the-art approaches in detecting haemorrhage.

Colorectal cancer is one of the deadliest cancer types worldwide. Therefore, an early detection is crucial to winning the fight against this disease. In this work, to help ease polyp detection, we present a fully convolutional network for colorectal polyp semantic segmentation (FCN-SEG4). This approach uses VGG16 as the backbone for feature extraction, followed by a series of transpose convolutions to get an accurate semantic segmentation. After training the model on the CVC-clinicDB dataset, an overall precision of 86.75% was reached. We trained FCN -SEG4 using other datasets to study the effect it may have on the results. This proposal proved good potential with room for improvement especially when it comes to performance speed.

Recently, pedestrian detection has attracted a lot of attention in recent years. It is known as a computer vision research hotspot, widely used in different fields. Despite the impressive progress of its approaches, their performance remains unsatisfactory. This paper proposes PedVis-Vgg-16a deep learning network for automatically detecting pedestrians presence in visible images. The suggested architecture is based on the fine-tuned VGG-16 architecture with modifications to the last block of the model. Different improvement components including data augmentation, parameter optimization, and parameter adaption, were taken to enhance the architecture performance. The newly designed architecture is validated on the publicly available dataset INRIA, which contains 4001 images and the results provided are satisfactory.

Wireless capsule endoscopy (WCE) is a non-surgical diagnostic procedure enabling the examination of the whole human gastrointestinal tract. Thus, a patient swallows a capsule that travels down the human digestive system and a camera captures wirelessly thousands of images that are transmitted to an external recording device. The diagnosis of these images need a specialist who can identify gastrointestinal abnormalities and it is very time-consuming. Recently, artificial intelligence and deep learning techniques aim to automate disease diagnosis and identi-fication of tumors in the gastrointestinal tract (GI) such as polyps, ulcers and bleeding, etc. In this paper, a deep learning method is proposed for gastrointestinal disease classification. The pre-trained model ResNetSO is fine-tuned through transfer learning to extract deep features from WCE images. The proposed algorithm is trained and tested on the publicly available dataset k-vasir capsule, which contains 14 different classes of gastrointestinal anomalies.

Abstract This paper addresses the problem of matching stereo images acquired by a stereo system mounted aboard an intelligent vehicle. The main idea behind the new method consists in involving temporal matching between a current stereo pair and its preceding one to achieve the spatial matching of the former stereo by involving the matching results obtained at the last frame. The proposed method is achieved in three main steps. First, an edge based disparity map is derived from the disparity map of the preceding frame, we call it the assisting disparity map (ADM). Second, for each scan‐line, a set of local ranges and global ranges are deduced from the ADM to keep only potential matching candidates. Third, the matching is done on the basis of dynamic programming algorithm by involving in the resulting local and global ranges we get from the last step. The proposed approach has been tested on both real and synthetic stereo sequences and the results demonstrate its effectiveness.

The human gastrointestinal (GI) tract may be infected by various diseases. If not detected at early stages, these abnormalities have the possibility to progress into gastric cancer, which is a common type of malignancies with yearly global cases exceeding one million. Endoscopy is a routinely used strategy for the examination of gastrointestinal tract diseases. During the examination, and due to many reasons like irregular morphologies, a huge number of frames, and exhaustion, gastrologists can miss some abnormalities. Thus, the automated classification of anomalies in endoscopic images is becoming necessary to assist medical diagnosis and reduce the cost and time of the medical process. Recent advances and high performance of deep learning techniques make it the best choice to adopt as a computer-aided-diagnosis strategy. In this paper, a novel deep learning model based deep convolutional neural network is proposed. Our model aims to automatically detect diseases from endoscopic images. The newly designed architecture is validated on the publicly available dataset KVASIR, which contains 8000 images. The results of our CNN approach compared to other well known pre-trained models showed important improvement and achieved 96.89% in terms of accuracy. The experiments demonstrated that the system can perform a high detection level without any human intervention.

Wireless capsule endoscopy (WCE) has revolutionised the diagnosis and treatment of gastrointestinal tract, especially the small intestine which is unreachable by traditional endoscopies. The drawback of the WCE is that it produces a large number of images to be inspected by the clinicians. Hence, the design of a computer‐aided diagnosis (CAD) system will have a great potential to help reduce the diagnosis time and improve the detection accuracy. To address this problem, the authors propose a CAD system for automatic detection of ulcer in WCE images. Firstly, they enhance the input images to be better exploited in the main steps of the proposed method. Afterward, segmentation using saliency map‐based texture and colour is applied to the WCE images in order to highlight ulcerous regions. Then, inspired by the existing feature extraction approaches, a new one has been proposed for the recognition of the segmented regions. Finally, a new recognition scheme is proposed based on hidden Markov model using the classification scores of the conventional methods (support vector machine, multilayer perceptron and random forest) as observations. Experimental results with two different datasets show that the proposed method gives promising results.

Wireless capsule endoscopy (WCE) proves its robustness as a great technology to examine the entire digestive tract or the small intestine. An automatic computer‐aided design method is proposed in this study, in a manner to differentiate between ulcer disease and normal WCE images. A multi‐scale analysis‐based grey‐level co‐occurrence matrix (GLCM) is conducted here. The main step, the co‐occurrence matrix (GLCM), is computed from each sub‐band Laplacian pyramid decomposition, so as to extract the common Haralick features. Moreover, the p ‐value and area under the curve are used to select the relevant characteristics from the feature descriptor. This proposed approach was separately applied to the components of CIELab colour space. Ulcer detection was performed using the support vector machine. The findings demonstrate an encouraging detection rate performance of 95.38% for accuracy and 97.42% for sensitivity based on the first dataset and an average accuracy of 99.25 and 98.51% of sensitivities for the second dataset.

In this paper, we present a new fast stereo matching method for omnidirectional images. The main contribution of the proposed approach is to involve the results obtained in the preceding frame in the matching of the current frame. The proposed approach consists of four steps. The first step is the spherical edge detection, which allows to compute the disparity only on significant features. The second step is the spherical rectification process, which is computed such that the epipolar lines coincide with the scan lines. In order to overcome the problem of blind spots near the epipoles caused by the distortions, we rectify three pairs of omnidirectional images: the current ones (spatial rectification), temporally adjacent left images, and temporally adjacent right images (temporal rectification).The third step is the stereo matching algorithm applied on the three rectified pairs. The final step is combining the three disparity maps into one single 360° disparity map. The proposed method tested in high-resolution omnidirectional images has shown to provide promising results towards stereo matching in omnidirectional imaging.

Wireless capsule endoscopy (WCE) is a contemporary method that can view the entire intestine. Because of its advantages, it has been widely used. However, physicians demand an automatic way to shorten the time required to analyze the produced images (over 55,000 images in one examination for one patient). In this paper a recognition system for CE imaging has been developed to automatically detect and recognize polyps. The proposed method is founded upon the extraction of Local Fractal Dimension (LFD) features over the detected SIFT Keypoints, these characteristics are then concatenated with Uniform Local Binary Pattern (LBPu) or Complete LBP (CLBP) so as to integrate the texture information. Various classifiers were used to distinguish the normal images from the polyp ones. Experimental results on a publicly available dataset demonstrate that the presented scheme can achieve 97.98% accuracy.

Traffic Sign Detection and Recognition is an important component of intelligent transportation systems. It has captured the attention of the computer vision community for several decades. In this paper, we propose a new traffic sign detection and recognition approach consisting of color segmentation, shape classification and recognition stages. In the first stage, the image is segmented using look-up tables and thresholding on the HSI color space. The second stage uses Distance to borders (DtBs) features and random forest classifier to detect circular, triangular and rectangular shapes among the segmented ROIs. The last stage consists in the recognition of the detected signs. It is performed using Random Forest as classifier and histogram of oriented gradients (HOG) together with local self-similarity (LSS) as features. Experimental results show that the proposed method achieves high recall, precision, and correct classification accuracy ratios, and is robust to various adverse situations.

In this paper, we present a fast approach for matching stereo images acquired by a stereo sensor embedded in a moving vehicle. The proposed approach exploits the disparity map already computed at the preceding frame to improve the matching results at the current one. An edge association method is used to track the edge curves over time. Local disparity constraints are computed for all the edge points that belong to the tracked edge curves. For the rest of edge points, we use a global disparity constraint, which is computed for each image line based on the preceding v-disparity. We integrate these constraints in the dynamic programming algorithm, which increases the matching results and speeds up the matching process.

This paper presents a LIDAR and vision-based pedestrian detection system. 3D LIDAR data are segmented into clusters. The clusters are projected on the corresponding image to get regions of interests (ROIs). Color-based HOG features are used together with SVM classifiers to detect pedestrian from extracted ROIs. The main contributions of the paper are the use of (LIDAR data to get ROIs and color-based HOG features for classification of ROIs. The proposed method has been tested on multisensors data and the results provided are satisfactory.

In this paper, we present a new fast method for matching stereo images acquired by a stereo sensor embedded in a moving vehicle. The method consists in exploiting the matching results obtained in one stereo pair (frame) for computing the disparity map of the following stereo pair. This can be achieved by finding a temporal relationship, which we named association, between consecutive frames. The disparity range of the current frame is deduced from the disparity map of the preceding frame and the association between the two frames. Dynamic programming technique is considered for matching the image features. The proposed approach is tested on virtual and real stereo image sequences and the results are satisfactory. The method is fast and able to provide about 20 millions disparity maps per second on a HP Pavilion dv6700 2.1GHZ.

This paper deals with a 2-D high resolution frequency estimation method dedicated to corrupted image by outliers. Outliers are particular data points that do not obey the assumed model. In this framework the well-known model of the sum of complex exponentials fails for the smallest fraction of a data set, which causes the classical estimators to produce inaccurate results. To alleviate this drawback, we propose a new robust iterative Levenberg-Marquardt (LM)-based method. The three main steps of the method we propose are as follows. First, we define a weight function based on the influence function which allows one to detect and correct the "wrong" data. The influence function measures the influence of a datum on the value of the parameter estimate. It is inspired from the so-called M-estimator. Second, a 2-D extension of the large sample approximation of the Maximum likelihood (ML) estimator is developed in order to estimate the image parameters. Third, the Levenberg-Marquardt (LM) technique is used to ensure the convergence of the ML estimator by performing the detection of "wrong" data for each iteration. The effectiveness of the proposed method is illustrated by some numerical simulations.