Image Segmentation Feature Research Articles

Improved segmentation of basal ganglia from MR images using convolutional neural network with crossover-typed skip connection.

Accurate and automatic segmentation of basal ganglia from magnetic resonance (MR) images is important for diagnosis and treatment of various brain disorders. However, the basal ganglia segmentation is a challenging task because of the class imbalance and the unclear boundaries among basal ganglia anatomical structures. Thus, we aim to present an encoder-decoder convolutional neural network (CNN)-based method for improved segmentation of basal ganglia by focusing on skip connections that determine the segmentation performance of encoder-decoder CNNs. We also aim to reveal the effect of skip connections on the segmentation of basal ganglia with unclear boundaries. We used the encoder-decoder CNNs with the following five patterns of skip connections: without skip connection, with full-resolution horizontal skip connection, with horizontal skip connections, with vertical skip connections, and with crossover-typed skip connections (the proposed method). We compared and evaluated the performance of the CNNs in the experiment of basal ganglia segmentation using T1-weighted MR brain images of 79 patients. The experimental results showed that the skip connections at each scale level help CNNs to acquire multi-scale image features, the vertical skip connections contribute on acquiring finer image features for segmentation of smaller anatomical structures with more blurred boundaries, and the crossover-typed skip connections, a combination of horizontal and vertical skip connections, provided better segmentation accuracy. This paper investigated the effect of skip connections on the basal ganglia segmentation and revealed the crossover-typed skip connections might be effective for improving the segmentation of basal ganglia with the class imbalance and the unclear boundaries.

Automated diagnosis of Retinopathy of prematurity from retinal images of preterm infants using hybrid deep learning techniques

SignificanceRetinopathy of prematurity (ROP) is a retinal disease in premature infants, which, when left untreated, can cause permanent blindness. Due to high variability and interobserver inconsistency in the diagnosis of ROP, there is a significant need to develop an automated system for the prediction of ROP. Although several methods have been used in automated diagnosis in ROP, no dedicated models have been described with adequate performance. Objectivesi) To develop a hybrid deep learning network for the prediction of ROP that can be used in the mass screening of infants. ii) To compare the performance of the proposed hybrid model with the Machine Learning classifiers and pre-trained CNN model. MethodsThe hybrid network is trained with 3200 and tested with 800 infant fundus images. Modified MultiResUNet and matched filter with first-order Gaussian derivative are used to segment the retinal vessels from the fundus images. The Gray level co-occurrence matrix (GLCM) and contour features of segmented images are extracted and selected using an embedded feature selection method. The selected features are evaluated using permutation importance and classified using the Random Forest classifier. ResultsThe proposed hybrid model predicts ROP with an accuracy of 94.5%, sensitivity of 94%, and specificity of 93%, surpassing the machine learning classifiers and the pre-trained models with reduced parameters and being computationally inexpensive. ConclusionsThe proposed hybrid model improves the quality of infant care by providing additional diagnostic assistance to clinicians. It also enhances the accessibility to remote healthcare centers for large-scale automated screening systems.

Ecommerce Website with Recommendation System Including Chatbot and Reverse Image Search

Abstract: In Today’s generation, it has become a need to get clear vision, prioritize and effortlessly to provide appropriate information to decrease the problem of information overload, which has created a major problem for many users using internet. Recommender systems is capable to solve this major problem by using searching technique through a huge volume of dynamically generated useful data to provide users with personalized content and services depending on their interest. Recommendation system plays an important role in the internet world and is used in many applications. Recommendation systems are of three types: 1) collaborative filtering 2) content based and 3) hybrid based approach. With the advancement of machine learning for decades, there are still numerous problems insolvable, such as image recognition and location detection, image classification, image generation, speech recognition, natural language processing and so on. In this emerging field of deep learning, the research on topic like image classification has always been the topmost traditional research to be evolved. Simultaneously, Image recognition technology is also beneficial to gradually respond in better way to the development of international indicators, and betterment the development and progress in various fields. Hence, image processing technology which is based on machine learning has always been widely used in feature image segmentation, classification and recognition, and is an important topic in various fields. It allows you to find similar products in your store in a very easy way. Let the customer find the product based on the photo. What's more, you no longer have to manually match similar products, it will be done automatically by AI algorithm. The main purpose of this project is to use the concept of a machine learning algorithm to build a Recommendation and Reverse Image Search and recognition with the help of convolutional neural networks (CNN).

A Novel Image Denoising and Segmentation Using Machine Learning with SRM Strategy

Objectives: In the last decade, artificial intelligence (AI) and machine learning (ML) have a significant impact on image analysis and segmentation. The most demanding aspect in digital image processing is efficient segmentation to extract the desired features or recognize the hidden patterns of the digital images, which are one of the major challenges. Even in existing segmentation techniques, due to under segmentation issues, particularly in remote or arial image analysis, necessary object features and residual details are not properly segmented. Therefore, the study presents a new method for image segmentation using machine learning, which addresses these limitations. Methods: This article presents a novel approach for arial image denoising and segmentation using unsupervised learning. The Discrete Wavelet Transform (DWT) is primarily used as a pre-processing tool, while edge details of the decomposed image are automatically preserved with the help of machine understanding. The use of a fast Non-Local (NL)-means filter provides a better visual effect for low-frequency image features, and the filtered image is partitioned into the number of clusters by Minimum- Spanning Tree (MST) clustering algorithm. Further, Statistical Region Merging (SRM) is used to eliminate the unwanted regions of the clustered image to give meaningful image details. Finally, the segmented image is projected at the wavelet restoration level. Here, the standard images are extracted from the SIPI database, and the system does not require training prototypes but works in an ”unsupervised” way. Findings: The proposed system segment 1014 regions in the first phase and the segmented image features efficiently recognize the hidden patterns of the arial image (7.1.06). However, to track the required object features, statistical Region Merging (SRM) is utilized, a drastic variation in the merging process eliminates the annoying details and brings effective outcomes, which involves 29 essential features with retained residual details at level-2 decomposition and is compared with leading segmentation techniques. The denoising performance of the suggested study provides a 21.73 Peak Signal to Noise Ratio (PSNR) value in 17.96 elapsed time. simulation results have proven to produce better segmentation and denoising with less iteration time. A qualitative evaluation through visual examination justifies the proposed study and is superior to some of the popular methods. Novelty/Applications: The utilization of wavelet transform to preserve edge features during segmentation is one of the key features of this method, and it significantly overcomes the under segmentation patterns. The combination of minimum-spanning tree clustering (MST) and statistical region merging (SRM) is the major strategic step in segmenting and detecting the essential object features of arial images. Keywords: DWT; Fast NLmeans filter; Wavelet denoising; MST; SRM

Hyper-fusion network for semi-automatic segmentation of skin lesions.

Segmentation of skin lesions is an important step for imaging-based clinical decision support systems. Automatic skin lesion segmentation methods based on fully convolutional networks (FCNs) are regarded as the state-of-the-art for accuracy. When there are, however, insufficient training data to cover all the variations in skin lesions, where lesions from different patients may have major differences in size/shape/texture, these methods failed to segment the lesions that have image characteristics, which are less common in the training datasets. FCN-based semi-automatic segmentation methods, which fuse user-inputs with high-level semantic image features derived from FCNs offer an ideal complement to overcome limitations of automatic segmentation methods. These semi-automatic methods rely on the automated state-of-the-art FCNs coupled with user-inputs for refinements, and therefore being able to tackle challenging skin lesions. However, there are a limited number of FCN-based semi-automatic segmentation methods and all these methods focused on 'early-fusion', where the first few convolutional layers are used to fuse image features and user-inputs and then derive fused image features for segmentation. For early-fusion based methods, because the user-input information can be lost after the first few convolutional layers, consequently, the user-input information will have limited guidance and constraint in segmenting the challenging skin lesions with inhomogeneous textures and fuzzy boundaries. Hence, in this work, we introduce a hyper-fusion network (HFN) to fuse the extracted user-inputs and image features over multiple stages. We separately extract complementary features which then allows for an iterative use of user-inputs along all the fusion stages to refine the segmentation. We evaluated our HFN on three well-established public benchmark datasets - ISBI Skin Lesion Challenge 2017, 2016 and PH2 - and our results show that the HFN is more accurate and generalizable than the state-of-the-art methods, in particular with challenging skin lesions.

A Meta-heuristic Approach for Design of Image Processing Based Model for Nitrosamine Identification in Red Meat Image

Background: Nitrosamine is a chemical, commonly used as a preservative in red meat whose intake can cause serious carcinogenic effects on human health. The identification of such malignant chemicals in foodstuffs is an ordeal. Objective: The objective of the proposed research work presents a meta-heuristic approach for nitrosamine detection in red meat using a computer vision-based non-destructive method. Methods: This paper presents an analytical approach for assessing the quality of meat samples upon storage (24, 48, 72 and 96 hours). A novel machine learning-based method involving the strategic selection of discriminatory features of segmented images has been proposed. The significant features were determined by finding p-values using the Mann-Whitney U test at a 95% confidence interval, which were classified using partial least square-discriminant analysis (PLS-DA) algorithm. Subsequently, the predicted model was evaluated by the bootstrap technique, which projects an outline for preservative identification in meat samples. Results: The simulation results of the proposed meta-heuristic computer vision-based model demonstrate improved performance in comparison to the existing methods. Some of the prevailing machine learning-based methods were analyzed and compared from a survey of recent patents with the proposed technique in order to affirm new findings. The performance of the PLS-DA model was quantified by the receiver operating characteristics (ROC) curve at all classification thresholds. A maximum of 100% sensitivity and 71.21% specificity was obtained from the optimum threshold of 0.5964. The concept of bootstrapping was used for evaluating the predicted model. Nitrosamine content in the meat samples was predicted with a 0.8375 correlation coefficient and 0.109 bootstrap error. Conclusion: The proposed method comprehends the double-cross validation technique, which makes it more comprehensive in discriminating between the edibility of foodstuff, which can certainly reinstate conventional methods and ameliorate existing computer-vision methods.

An integrated design of particle swarm optimization (PSO) with fusion of features for detection of brain tumor

Tumor in brain is a major cause of death in human beings. If not treated properly and timely, there is a high chance of it to become malignant. Therefore, brain tumor detection at an initial stage is a significant requirement. In this work, initially the skull is removed through brain surface extraction (BSE) method. The skull removed image is then fed to particle swarm optimization (PSO) to achieve better segmentation. In the next step, Local binary patterns (LBP) and deep features of segmented images are extracted and genetic algorithm (GA) is applied for best features selection. Finally, artificial neural network (ANN) and other classifiers are utilized to classify the tumor grades. The publicly available complex brain datasets such as RIDER and BRATS 2018 Challenge are utilized for evaluation of method and attained 99% maximum accuracy. The results are also compared with existing methods which evident that the presented technique provided improved outcomes which are clear proof of its effectiveness and novelty.

Automatic Detection of Tomato Leaf Deficiency and its Result of Disease Occurrence through Image Processing

Economic strength of a country is highly depends on agriculture productivity. More number of researches has been done based on detecting disease of plant by processing its leaf. However detecting disease after occurrence of it and identifying solution and source of it presence is a delay and useless process. The occurrence of disease in plants is due to absence of nutrient content in it. The main objective of this research work is to effectively detect deficiency in tomato leaf in order to protect it from disease occurrence. In our work, identification of deficiency in tomato leaf has been implemented using image processing technique. A plant grows in healthy way when it has its basic nutrients such as Nitrogen, phosphorus, potassium etc. Analyzing and detecting tomato plant leaf deficiency will identify occurrence of disease later. To detect deficiency in accurate way expectation and maximization segmentation process is implemented and features of segmented images have been extracted. Based on this extracted features classification is implemented to identify whether it is normal leaf or defected leaf. After identifying result the disease occurrence due to nutrient deficiency is shown. Therefore based on leaf image processing disease is identified by analyzing its deficiency in efficient way. Hence our research work prevents loss in tomato production and increases its growth and sales.

An Enhanced and Automatic Skin Cancer Detection using K-Mean and PSO Technique

Scientists have been trying to implement traditional methods around the world, particularly in developing countries, to reduce the death rate of skin cancer in humans. The scientific term is named as melanoma. But this effort always working hard as the system is costly, the low availability of experts and the conventional telemedicine. There are three types of skin cancer: basal cell cancer (BCC), squamous cell cancer, and melanoma. More than 90% of human is affected by ultraviolet (UV) radiation exposed to the sun. In this research, a skin cancer detection system (BCC) is designed in MATLAB. The images going to different processes such as Pre processing, feature extraction and classification. In pre-processing K-mean clustering is applied to determine the foreground and background of an image, since some part of background appear in the image after K-mean. Therefore, to resolve this problem Particle Swarm optimization (PSO) is applied. The segmented image features are extracted using Speed Up Robust Features (SURF), this helps to enhance the quality of the image. The Artificial neural network (ANN) is trained on the basis of these extracted features. To determine the efficiency of the system, the images are tested and performance parameters are measured. The detection accuracy determined by this model is about 98.7 5 is obtained.

SIR-DL: AN ARCHITECTURE OF SEMANTIC-BASED IMAGE RETRIEVAL USING DEEP LEARNING TECHNIQUE AND RDF TRIPLE LANGUAGE

The problem of finding and identifying semantics of images is applied in multimedia applications of many different fields such as Hospital Information System, Geographic Information System, Digital Library System, etc. In this paper, we propose the semantic-based image retrieval (SBIR) system based on the deep learning technique; this system is called as SIR-DL that generates visual semantics based on classifying image contents. At the same time we identify the semantics of similar images on Ontology, which describes semantics of visual features of images. Firstly, the color and spatial features of segmented images are we extracted and these visual feature vectors are trained on the deep neural network to obtain visual words vectors. The process of image retrieval is executed rely on semantic classification of SIR-DL according to the visual feature vector of the query image from which it produces a visual word vector. Then, we retrieve it on Ontology to provide the identities and the semantics of similar images corresponds to a similarity measure. In order to carry out SIR-DL, the algorithms and diagram of this image retrieval system are proposed after that we implement them on ImageCLEF@IAPR, which has 20,000 images. On the base of the experimental results, the effectiveness of our method is evaluated by the accuracy, precision, recall, and F-measure; these results are compared with some of works recently published on the same image dataset. It shows that SIR-DL effectively solves the problem of semantic-based image retrieval and can be used to build multimedia systems in many different fields.

BUILDING FOOTPRINT EXTRACTION FROM HIGH RESOLUTION SATELLITE IMAGERY USING SEGMENTATION

Abstract. Identification and mapping of urban features such as buildings and roads are an important task for cartographers and urban planners. High resolution satellite imagery supports the efficient extraction of manmade objects. For the planning and designing of Smart cities, building footprint information is an essential component, and geospatial technologies helps in creating this large mass of data inputs for designing and planning of smart cities. In this study segmentation approach is followed for building extraction. For extraction of buildings especially from the high resolution imagery, number of various semiautomatic and automatic methods have been developed till date to reduce the time and efforts required in manual building mapping. In this study, two semiautomatic image segmentation techniques are used for building extraction from high resolution imagery using algorithms- Multi-resolution segmentation and Rule based feature extraction, which are applied on Worldview 2 (2010) imagery of Dehradun area. The segmented image were further classified to extract buildings from the segmented image features. The study identify the usefulness of both the methods in building extraction and finds the optimum set of rules for extracting buildings from high resolution data sets. The True Positive Rate using Rule based feature extraction is 88.11 % compared to 85.46 % from Multi-resolution segmentation algorithm. The False Negative Rate (FNR) of Multi-resolution segmentation algorithm (16.5 %.) is very less compared to Rule based feature extraction (67.5 %). In the study the buildings were extracted with the accuracy of 88.9 %.

ANALYSIS OF IMAGE SEGMENTATION TECHNIQUES IN IMAGE PROCESSING

Image segmentation plays a very important role for many image video and computer vision applications. It is quite a relevant research area due to its wide usage in the field of medical, remote sensing and image retrieval. Image segmentation is used to identifying the objects as well as boundaries in the images. Based on the image feature image segmentation clusters or classifies the image into different parts. There are several algorithms proposed for segmenting an image prior to its recognition. This paper highlights the strength and a limitation of classification techniques applied to texture classification and reviews various algorithms like active contour model, fuzzy C means, fuzzy K means algorithm etc used in the segmentation process.

Leukaemia Detection in Microscopic Imagery using Optimization Algorithm

In this paper, automated approach of blood cancer detection is proposed. Usually microscopic images examined by experts manually are time consuming and less accuracy. The automated blood cancer detection system analyses the microscopic image and overcomes these drawbacks. The proposed system extracts the features of the image and applies filtering techniques. In this paper proposed method is cuckoo search optimization algorithm which is used in line with segmentation. The features of segmented image can be obtained from Scale invariant feature transform. Some of the features like PSNR, sensitivity, accuracy, etc. are calculated for leukemia detection. The performance is compared with the existing method ACO. The proposed system is tested on image dataset and 94.24% accuracy is achieved. The proposed system is successfully implemented in MATLAB.

SVM Classification of MRI Brain Images for Computer-Assisted Diagnosis

Magnetic Resonance Imaging is a powerful technique that helps in the diagnosis of various medical conditions. MRI Image pre-processing followed by detection of brain abnormalities, such as brain tumors, are considered in this work. These images are often corrupted by noise from various sources. The Discrete Wavelet Transforms (DWT) with details thresholding is used for efficient noise removal followed by edge detection and threshold segmentation of the denoised images. Segmented image features are then extracted using morphological operations. These features are finally used to train an improved Support Vector Machine classifier that uses a Gausssian radial basis function kernel. The performance of the classifier is evaluated and the results of the classification show that the proposed scheme accurately distinguishes normal brain images from the abnormal ones and benign lesions from malignant tumours. The accuracy of the classification is shown to be 100% which is superior to the results reported in the literature.

Histogram-based automatic thresholding for bruise detection of apples by structured-illumination reflectance imaging

Thresholding is an important step in the segmentation of image features, and the existing methods are not all effective when the image histogram exhibits a unimodal pattern, which is common in defect detection of fruit. This study was aimed at developing a general automatic thresholding methodology for fast and effective segmentation of bruises from the images acquired by structured-illumination reflectance imaging (SIRI). SIRI images, under sinusoidal patterns of illumination at a spatial frequency of 100 cycles m −1 , were acquired from 120 apple samples of four varieties with artificially created bruises and from another 40 apples with naturally occurred bruises. Subsequently, three sets of images, i.e., amplitude component (AC), direct component (DC) and ratio (i.e., dividing AC by DC), were derived from the original SIRI images. A unimodal thresholding method, called UNIMODE, was first applied to DC images for background removal, and then nine automatic thresholding techniques, including one unimodal and eight bimodal, were applied to the ratio images for bruise segmentation. It was found that severe over-segmentation occurred when using the bimodal thresholding methods, and this problem was mitigated by confining threshold selection to the lower part of the histogram that contained bruise information. Three bimodal thresholding techniques, i.e., INTERMODE (histogram valley emphasized), RIDLER (iterative thresholding), OTSU (clustering based) achieved the best bruise detection results with the overall accuracies of more than 90%. The overall detection results were further improved by integrating these techniques with the unimodal thresholding, due to reductions in the false positive error. The three bimodal thresholding techniques resulted in overall detection accuracies of 77–85% for naturally occurred bruises. This study has showed that the proposed automatic thresholding methodology provides a simple and effective tool for bruise detection of apples.

A semi-supervised fuzzy GrowCut algorithm to segment and classify regions of interest of mammographic images

According to the World Health Organization, breast cancer is the most common form of cancer in women. It is the second leading cause of death among women round the world, becoming the most fatal form of cancer. Despite the existence of several imaging techniques useful to aid at the diagnosis of breast cancer, x-ray mammography is still the most used and effective imaging technology. Consequently, mammographic image segmentation is a fundamental task to support image analysis and diagnosis, taking into account shape analysis of mammary lesions and their borders. However, mammogram segmentation is a very hard process, once it is highly dependent on the types of mammary tissues. The GrowCut algorithm is a relatively new method to perform general image segmentation based on the selection of just a few points inside and outside the region of interest, reaching good results at difficult segmentation cases when these points are correctly selected. In this work we present a new semi-supervised segmentation algorithm based on the modification of the GrowCut algorithm to perform automatic mammographic image segmentation once a region of interest is selected by a specialist. In our proposal, we used fuzzy Gaussian membership functions to modify the evolution rule of the original GrowCut algorithm, in order to estimate the uncertainty of a pixel being object or background. The main impact of the proposed method is the significant reduction of expert effort in the initialization of seed points of GrowCut to perform accurate segmentation, once it removes the need of selection of background seeds. Furthermore, the proposed method is robust to wrong seed positioning and can be extended to other seed based techniques. These characteristics have impact on expert and intelligent systems, once it helps to develop a segmentation method with lower required specialist knowledge, being robust and as efficient as state of the art techniques. We also constructed an automatic point selection process based on the simulated annealing optimization method, avoiding the need of human intervention. The proposed approach was qualitatively compared with other state-of-the-art segmentation techniques, considering the shape of segmented regions. In order to validate our proposal, we built an image classifier using a classical multilayer perceptron. We used Zernike moments to extract segmented image features. This analysis employed 685 mammograms from IRMA breast cancer database, using fat and fibroid tissues. Results show that the proposed technique could achieve a classification rate of 91.28% for fat tissues, evidencing the feasibility of our approach.

Performance Improvement of Plant Identification Model based on PSO Segmentation

Plant identification has been a challenging task for many researchers. Several researches proposed various techniques for plant identification based on leaves shape. However, image segmentation is an essential and critical part of analyzing the leaves images. This paper, proposed an efficient plant species identification model using the digital images of leaves. The proposed identification model adopts the particle swarm optimization for leaves images segmentation. Then, feature selection process using information gain and discritization process are applied to the segmented image's features. The proposed model was evaluated on the Flavia dataset. Experimental results on different kind of classifiers show an improvement in the identification accuracy up to 98.7%. Index Terms—Plant identification; Segmentation; Particle Swarm Optimization; Information Gain; Discretization.

Assessment of image features for vessel wall segmentation in intravascular ultrasound images.

Intravascular ultrasound (IVUS) provides axial greyscale images, allowing the assessment of the vessel wall and the surrounding tissues. Several studies have described automatic segmentation of the luminal boundary and the media-adventitia interface by means of different image features. The aim of the present study is to evaluate the capability of some of the most relevant state-of-the-art image features for segmenting IVUS images. The study is focused on Volcano 20MHz frames not containing plaque or containing fibrotic plaques, and, in principle, it could not be applied to frames containing shadows, calcified plaques, bifurcations and side vessels. Several image filters, textural descriptors, edge detectors, noise and spatial measures were taken into account. The assessment is based on classification techniques previously used for IVUS segmentation, assigning to each pixel a continuous likelihood value obtained using support vector machines (SVMs). To retrieve relevant features, sequential feature selection was performed guided by the area under the precision-recall curve (AUC-PR). Subsets of relevant image features for lumen, plaque and surrounding tissues characterization were obtained, and SVMs trained with these features were able to accurately identify those regions. The experimental results were evaluated with respect to ground truth segmentations from a publicly available dataset, reaching values of AUC-PR up to 0.97 and Jaccard index close to 0.85. Noise-reduction filters and Haralick's textural features denoted their relevance to identify lumen and background. Laws' textural features, local binary patterns, Gabor filters and edge detectors had less relevance in the selection process.

Describing Images using a Multilayer Framework based on Qualitative Spatial Models

To date most research in image processing has been based on quantitative representations of image features using pixel values, however, humans often use and semantic knowledge to describe and analyze images. To enhance cognitive adequacy and tractability, we here present a multilayer framework based on qualitative spatial models. The layout features of segmented images are defined by qualitative spatial models which we introduce, and represented as a set of qualitative spatial constraints. Assigned different semantic and context knowledge, the image segments and the qualitative spatial constraints are interpreted from different perspectives. Finally, the knowledge layer of the framework enables us to describe the image in a natural way by integrating the domain-specified semantic constraints and the spatial constraints.

Analyzing the Influences of Camera Warm-Up Effects on Image Acquisition

This article presents an investigation of the impact of camera warm-up on the image acquisition process and therefore on the accuracy of segmented image features. Based on an experimental study we show that the camera image is shifted to an extent of some tenth of a pixel after camera start-up. The drift correlates with the temperature of the sensor board and stops when the camera reaches its thermal equilibrium. A further study of the observed image flow shows that it originates from a slight displacement of the image sensor due to thermal expansion of the mechanical components of the camera. This sensor displacement can be modeled using standard methods of projective geometry in addition with bi-exponential decay terms to model the temporal dependency. The parameters of the proposed model can be calibrated and then used to compensate warm-up effects. Further experimental studies show that our method is applicable to different types of cameras and that the warm-up behaviour is characteristic for a specific camera.