Image segmentation is a key topic in image processing and computer vision with applications such as scene understanding, medical image analysis, robotic perception, video surveillance, augmented reality, and image compression, among many others. Various algorithms for image segmentation have been developed in the literature. Recently, due to the success of deep learning models in a wide range of vision applications, there has been a substantial amount of works aimed at developing image segmentation approaches using deep learning models. In this survey, we provide a comprehensive review of the literature at the time of this writing, covering a broad spectrum of pioneering works for semantic and instance-level segmentation, including fully convolutional pixel-labeling networks, encoder-decoder architectures, multi-scale and pyramid based approaches, recurrent networks, visual attention models, and generative models in adversarial settings. We investigate the similarity, strengths and challenges of these deep learning models, examine the most widely used datasets, report performances, and discuss promising future research directions in this area.

Image Segmentation Using Deep Learning: A Survey

A review on brain tumor segmentation of MRI images.

Currently, we witness the severe spread of the pandemic of the new Corona virus, COVID-19, which causes dangerous symptoms to humans and animals, its complications may lead to death. Although convolutional neural networks (CNNs) is considered the current state-of-the-art image classification technique, it needs massive computational cost for deployment and training. In this paper, we propose an improved hybrid classification approach for COVID-19 images by combining the strengths of CNNs (using a powerful architecture called Inception) to extract features and a swarm-based feature selection algorithm (Marine Predators Algorithm) to select the most relevant features. A combination of fractional-order and marine predators algorithm (FO-MPA) is considered an integration among a robust tool in mathematics named fractional-order calculus (FO). The proposed approach was evaluated on two public COVID-19 X-ray datasets which achieves both high performance and reduction of computational complexity. The two datasets consist of X-ray COVID-19 images by international Cardiothoracic radiologist, researchers and others published on Kaggle. The proposed approach selected successfully 130 and 86 out of 51 K features extracted by inception from dataset 1 and dataset 2, while improving classification accuracy at the same time. The results are the best achieved on these datasets when compared to a set of recent feature selection algorithms. By achieving 98.7%, 98.2% and 99.6%, 99% of classification accuracy and F-Score for dataset 1 and dataset 2, respectively, the proposed approach outperforms several CNNs and all recent works on COVID-19 images.

/pdf/covid-19-image-classification-using-deep-features-and-3so55ekumz.pdf

COVID-19 image classification using deep features and fractional-order marine predators algorithm.

A comprehensive survey of clustering algorithms: State-of-the-art machine learning applications, taxonomy, challenges, and future research prospects

To improve the recognition model accuracy of crop disease leaves and locating diseased leaves, this paper proposes an improved Faster RCNN to detect healthy tomato leaves and four diseases: powdery mildew, blight, leaf mold fungus and ToMV. First, we use a depth residual network to replace VGG16 for image feature extraction so we can obtain deeper disease features. Second, the k-means clustering algorithm is used to cluster the bounding boxes. We improve the anchoring according to the clustering results. The improved anchor frame tends toward the real bounding box of the dataset. Finally, we carry out a k-means experiment with three kinds of different feature extraction networks. The experimental results show that the improved method for crop leaf disease detection had 2.71% higher recognition accuracy and a faster detection speed than the original Faster RCNN.

/pdf/deep-learning-based-object-detection-improvement-for-tomato-4pfas6p35o.pdf

Deep Learning-Based Object Detection Improvement for Tomato Disease

Image Segmentation Using K -means Clustering Algorithm and Subtractive Clustering Algorithm

Image segmentation has been considered as the first step in the image processing. An efficient segmentation result would make it easier for further analysis of image processing. However, there exits many algorithms and approaches for image segmentation. Clustering is one of the commonly used image segmentation techniques. In this paper, we have briefly describe some of the clustering techniques and discuss some of the recent works by researchers on these techniques.

A Survey on Image Segmentation Methods using Clustering Techniques

Image segmentation has considered an important step in image processing. Fuzzy c-means (FCM) is one of the commonly used clustering algorithms because of its simplicity and effectiveness. However, FCM has the disadvantages of sensitivity to initial values, falling easily into local optimal solution and sensitivity to noise. To tackle these disadvantages, many optimization-based fuzzy clustering methods have been proposed in the literature survey. Particle swarm optimization (PSO) has good global optimization capability and a hybrid of FCM and PSO have improved accuracy over tradition FCM clustering. In this paper, a new image segmentation method based on Dynamic Particle swarm optimization (DPSO) and FCM algorithm along with the noise reduction mechanism is proposed. DPSO has the advantages to change the inertia weight and learning parameters dynamically. It adopts the inertia weight according to the fitness value and learning parameters along with time. The proposed method combines DPSO with FCM, using the advantages of global optimization searching and parallel computing of DPSO to find a superior result of the FCM algorithm. Moreover, a noise reduction mechanism based on the surrounding pixels is used for enhancing the anti-noise ability. The synthetic image and Magnetic Resonance Imaging (MRI) have been used for testing the proposed method by introducing different types of noises and the results show that the proposed algorithm has better performance and less sensitive to noise.

An image segmentation approach based on fuzzy c-means and dynamic particle swarm optimization algorithm

Pulmonary nodule detection is an image processing approach to detect lung cancer from thoracic computed tomography (CT) images Lung segmentation is the first step to segment lung regions from CT images Nodule candidates are detected from the segmented lung regions Further, nodule classification is performed to identify the true nodules from the false positives In a recent paper, linear discriminant analysis (LDA) shows better performance than quadratic discriminant analysis for nodule classification In the proposed method, the performance of the existing LDA method is improved by introducing additional discriminant features extracted by using multiple discriminant analysis Further, a noble nonlinear classifier is used to overcome the limitation of the linear classifier For the nonlinear classification, multilayer perceptron is used using a new powerful learning algorithm The new learning algorithm is a combination of genetic algorithm (GA) and particle swarm optimization Finally, the performance of the proposed method is compared with the existing LDA and convolutional neural network methods The new learning algorithm of the proposed method is also compared with backpropagation (BP) and GA optimized BP algorithms The overall performance of the proposed method is remarkable

Pulmonary nodule detection on computed tomography using neuro-evolutionary scheme

Speckle, a multiplicative noise, is an inherent property of ultrasound imaging It reduces the contrast and resolution of the ultrasound images Thus, it creates a negative effect on image interpretation and diagnostic tasks In this paper, a modified non-local means filter using Bhattacharyya distance is proposed In the non-local mean, noise free pixel is estimated as a weighted mean of image pixels, where weights are calculated according to the similarity between image patches Similarity between the patches is measured by comparing pixel intensities In this work, instead of comparing pixel intensities for measuring similarities, blocks are used for measuring similarities based on Bhattacharyya distance Quantitative results on both simulated and real ultrasound images show the effectiveness of the proposed method compared to other well-known methods

Yambem Jina Chanu

Papers

Image Segmentation Using K -means Clustering Algorithm and Subtractive Clustering Algorithm

A Survey on Image Segmentation Methods using Clustering Techniques

An image segmentation approach based on fuzzy c-means and dynamic particle swarm optimization algorithm

Pulmonary nodule detection on computed tomography using neuro-evolutionary scheme

Speckle reduction of ultrasound medical images using Bhattacharyya distance in modified non-local mean filter