Machine Learning (ML) is a subfield of Artificial Intelligence (AI) that is increasingly applied to several medical diagnosis tasks, including a wide range of diseases. Importantly, various ML models were developed to address the complexity of Parkinson's Disease (PD) diagnosis. PD is a neurodegenerative disease characterized by motor and non-motor disorders where its syndromes affect the daily lives of patients. Several Computer Aided Diagnosis and Detection (CADD) systems based on hand-crafted ML algorithms achieved promising results in distinguishing PD patients from Healthy Control (HC) subjects and other Parkinsonian syndrome categories using clinical data (e.g., speech and gait impairments) and medical imaging [e.g., Position Emission Tomography (PET) and Single Photon Emission Computed Tomography (SPECT)]. Despite the good performance of hand-crafted ML algorithms, there is still a problem linked to the features' extraction and selection. In fact, Deep Learning DL has provided an ultimate solution for the features' extraction and selection related issue. An important number of studies on the diagnosis of PD using DL algorithms were developed recently. This study provides an overview of the application of hand-crafted ML algorithms and DL techniques for PD diagnosis. It also introduces key concepts for understanding the application of ML methods to diagnose PD.

This study presents a comprehensive survey on mixed impulse and Gaussian denoising filters which are applied to an image in order to gauge the effects of this type of noise combination and to then determine optimal ways that can overcome such effects. The random noise model considered in this survey is the combined effect of impulse (salt and pepper) and Gaussian noise. After describing the noise models, the denoising filters which are applied to the images are classified and explained according to their design structure, the type of filters they use, the noise level they could overcome, and the limitations they face. This survey covers all related denoising methods and provides an assessment of the strengths and practical limitations of the different classes of denoising filters.

To reduce the computation required in determining the proper scale of salient object, a fast visual saliency based on multi-scale difference of Gaussians fusion in frequency domain (MDF) is proposed. First, based on the phenomenon that the foreground energy is highlighted and densely distributes on certain band of spectrum, the scale coefficients of foreground in an image can be literately approximated on the amplitude spectrum. Next, relying on the linear integration property of Fourier transform, the feature spectrum is obtained through the weighted infinite integral of difference of Gaussian feature maps with respect to the scale of object. Then, the saliency of each channel is obtained from feature spectrum by the inverse Fourier transform and scale filtering. Finally, through the channel integration, the MDF saliency map is obtained. Experiments on Li-Jian data set demonstrate that combined with most appropriate colour space and scale filter, MDF achieves obvious acceleration (5.4 times faster than frequency domain analysis and spatial information) while getting desired accuracy (area under the curve, 0.8814 at Li-Jian data set), which achieves the best accuracy efficiency trade-off.

Multi-source domain adaptation (MDA) aims to leverage knowledge from multiple source domains to improve the classification performance on target domains. Different degrees of distribution discrepancies between every two domains pose a huge challenge to MDA tasks. Most works focus on extracting features shared by all domains, which is critical but not enough to reduce distribution discrepancies. In this paper, we propose a method named as common-specific feature learning (CSFL). Constituting a framework of feature learning, CSFL explores a subspace where the combination of common and specific features makes learned representations comprehensive. Based on this framework, we conduct a metric learning method for learning a discriminative feature representation. Considering redundant information caused by source domains is likely to hurt the performance, we impose an effective low-rank constraint to remove the redundant information. Further, we adopt structure consistent constraint to preserve the local structure in each domain. CSFL has obtained about 1–5% improvement of mean accuracy, compared to the state-of-the-art shallow methods. Further, compared with 90.2% and 89.4% of the best baseline deep method, CSFL achieves mean accuracy of 90.8% and 89.7% on the Office-31 and ImageCLEF-DA datasets respectively. The encouraging results validate the effectiveness of our method.

The novel coronavirus has spread quite rapidly across the globe. The current testing rate is failing to match the exponential rate of rising cases. Moreover, the available testing methodologies are expensive and time-consuming. A sensitive automated diagnosis is one of the biggest need of the hour. In the proposed work, the authors analyse the chest X-ray images of normal, pneumonia and coronavirus disease-2019 (COVID-19) patients and process them to boost the COVID-specific features (opacities etc.), which enable to perform sensitive identification of COVID-19 patients. The sets of original and processed images are used with a stack of pre-trained deep models for ensemble learning. They used VGG-16 as base-learners, trained with a diverse set of inputs followed by a logistic regression model, the meta learner, to combine the base-learner predictions. The proposed fusion-based model is trained and tested for three types of classification, TYPE-I: binary (NORMAL/ABNORMAL), TYPE-II: binary (PNEUMONIA/COVID-19) and TYPE-III: multi-class (NORMAL/PNEUMONIA/COVID-19). The diagnosis results are quite promising, with high accuracy and sensitivity values for all the cases. The proposed algorithm can be used to assist the medical experts for quick identification and isolation of COVID-19 patients and thereby mitigating the effect of the virus.

Diabetic Maculopathy, the major cause of vision loss, occurs due to uncontrolled diabetes. This affects the retinal layers of the eye causing bleeding of vessels, which results in Diabetic Macular Edema (DME), macular detachment etc., Basically three structural changes are involved in DME viz., Cystoid Macular Edema (CME), Serous Macular Detachment (SMD) and Intra Retinal Fluid (IRF). These changes may also coexist with each other such as CME with SMD or CME with IRF etc. In this work, the retinal images acquired through Spectral Domain Optical Coherence Tomography (SDOCT) imaging modality, which would provide high level of precision and resolution of the retinal layers are utilized. An automated algorithm to detect and differentiate seven types of DME based on deep learning is implemented using transfer learning on three Convolutional Neural Networks (CNNs), ResNet 50, VGGNet and AlexNet. In detection of DME, ResNet 50 performs excellently well, when compared with Alexnet and VGGNet, because of its depth and skip connection features. The average values of the statistical parameters such as accuracy (0.993), F1 score (0.975), Mathews Correlation Coefficient(MCC)(0.972) of ResNet are high when compared to that of AlexNet and VGGNet

The success of clustering algorithms in object segmentation depends on the quality of the evaluation criterion. However, sonar images are seriously affected by noise. Most of the existing evaluation criteria such as the Davies Bouldin (DB) criterion only considers their pixel value, and sonar image information extraction is not sufficient. As a result, they fail to achieve good underwater object segmentation results. To overcome this problem, this study proposes an improved DB criterion with pulse -coupled neural network (PCNN), which is called the DB-PCNN. In the calculation process of DB-PCNN, the role of internal activity items in PCNN is considered, which can make better use of pixel information in adjacent space on the sonar image. The experimental results show that DB-PCNN can further improve the accuracy of underwater object segmentation and has certain adaptability to different optimisation frameworks.

This study proposes a hybrid feature convolutional neural network (HFCNN) model for the complete description of three-dimensional (3D) point cloud features. The HFCNN confers sensitivity to the local, global, and single-point properties simultaneously by a feature vector space expansion. Wherein, a pointwise convolutional network sub-model realises the extraction of the local features by using a pointwise convolutional operator to process point cloud data directly. To consider the global properties of the point cloud, a central-point radiation model is constructed as an input of the feature layer in a non-network form. Meanwhile, the single-point behaviour is characterised by the solo point coordinate information in the network. Within the constructed solo-local-global feature space, i.e. the fusion of single point feature, local feature and global feature, the HFCNN model can handle 3D point cloud data with unstructured and unordered properties. The HFCNN can be directly applied to the point cloud classification and segmentation without the modification of the CNN structure and training procedure. The experimental results have shown the effectiveness of the proposed model in prediction of class labels and point-by-point labels.

Scene flow estimation with RGB-D frames is receiving increasing attention in digital video processing and computer vision due to the widespread use of depth sensors. Existing methods based on object segmentation have shown their effectiveness for object occlusion and large displacement. However, improper segmentation often causes incomplete segmented areas or incorrect edges, which will result in inaccurate occlusion inference and scene flow estimation. To this end, an automatic layered RGB-D scene flow estimation method is proposed, which achieves more accurate layering of objects in depth image by exploring motion information. The authors exploit super-pixel segmentation for initial layering, which is beneficial for preserving edges and integrity of the objects. Furthermore, an optical flow which is highly correlated with the scene is also used for automatic layering. Finally, the utilisation of super-pixel segmentation and motion information would ensure the integrity of the object area and improve the accuracy of scene flow. They have validated their approach both qualitatively and quantitatively on several public datasets. Experimental results show that the proposed method is able to preserve the integrity of the object and achieve lower root mean square error and average angular error as compared with the current state-of-the-art algorithms.

In this study, the authors are interested in building a unified deep learning framework to solve the recognition problem of both single-line and double-line car license plates. Most existing methods are designed for single-line license plate recognition. For double-line cases, detection and segmentation are usually adopted firstly to locate each line of characters. These methods are usually environmentally sensitive and will bring propagation of error between segmentation and recognition. To solve the problems, the authors propose a unified method that can recognise both single-line and double-line license plates in an end-to-end way without line segmentation and character segmentation. Specifically, an improved lightweight convolutional neural network is used to extract features efficiently. Then, the multi-task learning strategy is used to simultaneously perform license plate classification and character recognition. Finally, recognition task is treated as sequence labelling problems, which are solved by connectionist temporal classification directly. Experimental results indicate that the proposed method significantly outperforms the previous state-of-the-art methods on both public datasets and the synthetic dataset.

This study presents the implementation of image kernels used for filtering and enhancing the images using reversible logic gates, a first in reversible logic literature. Image enhancement/filtering is achieved by performing convolution of an image with a filter kernel. This work proposes reversible logic based design and implementation of six filter kernels. The filter kernels implemented are Gaussian blur, Laplacian outline, Sobel, Emboss, Sharpen and Prewitt edge detection. The kernels are implemented individually using reversible logic gates and the designs are measured in terms of quantum cost, garbage outputs, ancilla inputs and gate count. The functional verification is carried out using standard images on Kintex 7 FPGA platform. The filtered images from the proposed design have an average structural similarity index of 0.92.

Nowadays, skin cancer is one of the most dangerous forms of cancer found in humans. There are various types of skin cancer, like basal, melanoma, carcinoma, and the squamous cell from which the melanoma is unpredictable. Thus, skin cancer detection in the early stage is very useful to treat it successfully. Hence, this study introduces a new algorithm called social bat optimisation algorithm for skin cancer detection. Initially, the pre-processing is done for the input image to eliminate the noise and artefacts present in the image. Then, the pre-processed image is fed to the feature extraction step where the features are extracted based on convolutional neural network features, and the local pixel pattern-based texture feature (local PPBTF). Here, the PPBTF is the combination of texture features and pixel pattern-based features in which the equation of PPBTF is modified based on the local binary pattern. Subsequently, the classification is done based on the extracted features using a deep stacked auto-encoder, which is trained by the proposed social bat optimisation. The performance of skin cancer detection based on the proposed model is evaluated based on accuracy, sensitivity, and specificity. The proposed model achieves the maximal accuracy of 93.38%, maximal sensitivity of 95%, and the maximal specificity of 96% for K-fold.

In this study, an unsupervised clustering algorithm is proposed to label superpixel density images. Firstly, the authors propose a novel superpixel segmentation algorithm driven by a modified fuzzy C-means objective function, Kullback–Leibler (KL) divergence, and an entropy term, which generate superpixels with good boundary adherence and intensity homogeneity. In this model, the logarithm of Gaussian distribution as a new distance metric is used to improve the accuracy of boundary pixel classification, the KL divergence is applied to regularise the fuzzy objective function. Based on this model, the generated superpixel intensity images with a highly distinctive background colour from the colour of the target are obtained. Grouping cues generated by superpixels can affect the performance of image clustering greatly. Next, according to the small amount of clustering data generated by the superpixel intensity images, they construct a non-symmetric mixture model based on a mixture of Gaussian distribution and Cauchy distribution for implementing image clustering. Thus, clustering of colour images is transformed into clustering of these newly generated data. The advantage of this model is its well adaption to different shapes of observed data. Experimental results on publicly available data sets are provided to demonstrate the effectiveness of the proposed algorithm.

Since histopathological images exist in various forms, performing segmentation on these images is tedious. While in cancer-free colon tissue, epithelial cells generally have an elliptical shape; their structure alters in a malignant tissue. This study proposes a technique consisting of colon biopsy image segmentation and a hybrid set of features for classification, and is evaluated on multiple databases with various levels of magnifications. This study presents a novel image segmentation method with multi-level thresholding based on Rényi's two-dimensional entropy with a cultural algorithm (2DRCA). Based on the entropy, elliptical epithelial cells, being the region of interest, are identified from the segmented background. After successful segmentation, shape descriptors are extracted with morphological operations. Two sets of texture features (grey-level co-occurrence matrix and block-wise elliptical local binary pattern) are calculated based on pre-processed grey-scale colon images. The proposed hybrid feature vector set, then concatenates the extracted features for training and testing with a random forest classifier. The proposed segmentation and classification model is evaluated by considering four data sets consisting of various colon images at different magnifications. In addition, it is evaluated by multiple performance measures and compared with existing techniques.

A new denoising technique is proposed in this study for synthetic aperture radar (SAR) images corrupted by speckle noise. The authors method extract informative features from a noisy speckled image, and then a denoised version of this image is estimated from the informative gradients, which are restricted to the features of this image. The technique of extracting features is designed on the framework of multifractal formalism followed by a reconstruction technique for the informative gradients based on the total variational (TV) regularisation framework. Experimental results demonstrate that the proposed approach is able to retain the finer details of the original image while removing noise. The superiority of the proposed approach is manifested qualitatively and quantitatively on comparing with state-of-the-art denoising techniques.

Video forensics is one of the hot topics in multimedia forensics. Nowadays, spreading fake videos across the internet is becoming a profession mainly in politics and entertainment. In this study, a novel two-stage inter-frame video forgery detection technique is proposed. The first stage analyses spatio-temporal feature flow consistency to detect suspicious tamper points. Identifying the type of forgery and validating the recovered video are done in the second stage. Earth mover's distance is used as a similarity metric in both stages. The authors concentrate on a robust inter-frame forgery detection approach which can be applied for any challenging video. Compression at a higher rate, noise addition, and filtering are the anti-forensic tricks used by forgers to fool forensic techniques. However, most of the literature in video forgery detection has handled these issues as post-processing attacks and reported lesser accuracies for it. Hence the authors propose a robust and efficient forgery detection technique capable of identifying all kinds of inter-frame forgeries in videos. Experimental evaluation of the public video data set shows that the proposed approach outperforms existing approaches with an improved rate of robustness.

Apple fruit sorting has been an important postharvest process carried on for the sorting of diseased apple fruits. A fuzzy cluster-based thresholding (FCBT) method for segmenting the region of interest from an apple image has been proposed for sorting apples in this study. As the first step, the acquired RGB colour image of an apple fruit was converted into a greyscale image. Then, five different fuzzy cluster bins with overlapped pixel ranges were taken and greypixel values were binned into them. A cluster with the maximum number of pixels was selected for calculating the threshold value. The region of interest from the apple image was then segmented using the proposed FCBT value. Features extracted from the segmented images were given as input to a fully-connected deep neural network for a classification. The performance of the FCBT method was compared with similar greyscale thresholding methods like Otsu's and Kapur's methods. The visual segmentation accuracy and the execution speed showed that the FCBT outperformed the other methods in segmenting the diseased area. A fully-connected deep neural network model with the FCBT image extracted features as input values gave a 98.33% accuracy rate in sorting the apple images.

Underwater images often undergo distortions from scattering, absorption, colour loss, diffraction, polarisation, and varying attenuation depending on the light frequency, due to the water medium. This study answers problems associated with the recovery of underwater images, by developing novel depth map estimation which may be used as an intermediary step for underwater image restoration. The depth map is an important factor for the recovery of the underwater image, as it has been shown that proper estimation of depth, results in better restoration of the underwater image. The proposed algorithm estimates a depth map from a single view image, using blurriness and lighting information, obtained using a simple background neutralisation method. Results from the algorithm on selected raw underwater image data set have been compared with other algorithms and it has been shown that the proposed method gives Pearson coefficients in the range of 0.75–0.95, which are considerably higher and better log root mean square value compared to other methods in the literature. This suggests a better estimation of depth by the proposed method.

Image segmentation is one of the most important aspects of image processing that has many applications in image analysis, computer vision, and machine vision. So far, researchers have proposed many image segmentation algorithms that provide a relatively appropriate response for a particular application and over a specific type of input images. At the same time, they have not yet suggested a simple, fast, and efficient segmentation method that can apply to a wide range of natural and texture images. In this study, the authors present a new adaptive method for selecting Gabor filters using the image frequency content. Adaptive spectral features of the image are then extracted using only these selected filters from a Gabor filter bank. A pre-setting parameter Gabor filter is used to extract the gradient of each feature in x and y directions. Combining these features gradient, the texture gradient image is calculated, and then by selected markers for different texture regions and applying the marker-based watershed transform on texture gradient image, the authors present a new image segmentation method that has a less computational cost and yields better results compared to using all other filter banks.

Image fusion is essential to produce a complementary and comprehensive image, with the source images derived from different sensors, captured from different illumination conditions. In this study, a novel multi-scale image fusion based on the combination of non-subsampled contourlet transform (NSCT) and rolling-guidance filter (RGF) is used to enhance the edges and texture details better than the conventional methods. Initially, infrared (IR) and visible (VIS) source images are multi-scale decomposed to low-frequency and high-frequency sub-band coefficients by NSCT for the best representation of edges and curves. Further, the low-frequency coefficients are decomposed into the base and detail layers by a combination of RGF and GF (Gaussian filter) to retain the features in multiple scales and to reduce halos near the edges. Base layers are fused by saliencybased fusion rule and detail layers are fused by Max absolute rule. High-frequency coefficients are fused by consistency verification based fusion rule to preserve visual details and to suppress noise from source images. Finally, the image is reconstructed by inverse NSCT with good visual perception. Experimental results are evaluated by different evaluation metrics and the results suggest that the proposed method results with better improved source information, clarity and contrast.

The minutiae set defined by the ISO/IEC 19794-2 is one of the prevalent feature used in fingerprint recognition systems. Unfortunately, such characteristic of unordered and variable-sized minutiae information causes a restriction on the operation in some advanced template protection methods (e.g. fuzzy commitment), which usually require an ordered and fixed-length binary feature representation as the system input. In this study, in order to simultaneously extend the application of fingerprint recognition and provide satisfactory system performance, the authors propose a novel fixed-length bit-string conversion framework based on spectral clustering and the proposed newly designed discriminative fingerprint representation called minutia vicinity combined feature (MVCF). The proposed method consists of three stages: (i) the extraction of MVCF, (ii) bit conversion via the spectral clustering algorithm, and (iii) matching. Benefiting from feature invariance, fixed-length and bit-oriented coding, merits such as fast matching and decent accuracy are well guaranteed. The performance evaluation is conducted on six publicly available benchmark data sets: FVC2002 DB1, DB2, DB3 and FVC2004 DB1, DB2, DB3 confirms the superiority of the proposed method and suggests the promise of migrating to some other domains (e.g., template protection).

In this study, a novel reversible data hiding (RDH) in encrypted domain scheme for coding channel based on sliding-block segmentation and adaptive steganography is proposed. The proposed scheme enriches the residual information with as little additional encryption information as possible to improve the testing error rate of a steganalyser by sliding-block segmentation with bit stream encryption. The specific encryption process effectively weakens the correlation between the adjacent pixels and minimises the size of key stream bits. The encryption key can be further embedded in the channel code stream before transmitted in the channel. Experimental analysis shows that the image encrypted by the proposed RDH scheme can achieve a peak-signal-to-noise ratio of >50 dB, as the payload is 0.5 bits per pixel (bpp). In terms of security performance, compared with the state-of-the-art methods, their method has a higher testing error rate when the steganalyser is utilised. Even if the payload is 0.5 bpp, the testing error rate is >0.25.

Aiming at the problems of colour distortion and residual dehazing in the existing image dehazing methods when processing outdoor images, a method of image dehazing based on the cycle generative adversarial network is proposed. Taking the cycle generative adversarial network as the overall framework of the model, firstly, the neural network is trained to obtain the mapping relationship between haze images and haze-free images. Secondly, to accelerate the convergence speed of network training, using residual structure to improve network stability and reduce parameters. Then, a new loss function is proposed, fusing the Wasserstein distance into adversarial loss and cycle consistency loss, which reduces the deviation between the generated dehaze image and the real haze-free image, and alleviating the problems of colour distortion and haze removal residue. Finally, use the optimised bounded ReLU (BReLU) activation function instead of the original activation function to improve the transmittance reflecting the haze depth information. The experimental results demonstrated that, compared with the comparison method, the proposed method improves the peak signal-to-noise ratio, structure similarity, information entropy, and average gradient, and has an achieved better performance in dehazing.

Prediction methods of glioblastoma tumours growth constitute a hard task due to the lack of medical data, which is mostly related to the patients' privacy, the cost of collecting a large medical data set, and the availability of related notations by experts.In this study, the authors propose a synthetic medical image generator (SMIG) with the purpose of generating synthetic data based on the generative adversarial network in order to provide anonymised data. In addition, to predict the glioblastoma multiform tumour growth the authors developed a tumour growth predictor based on end to end convolution neural network architecture that allows training on a public data set from the cancer imaging archive (TCIA), combined with the generated synthetic data. The authors also highlighted the impact of implicating a synthetic data generated using SMIG as a data augmentation tool. Despite small data size provided by TCIA data set, the obtained results demonstrate valuable tumour growth prediction accuracy.

The wavelet block tree coding (WBTC) algorithm is an efficient wavelet-based image coder. In this coder, multiple spatial orientation trees are combined together to make a single block tree. It uses three ordered lists to keep track of significant/insignificant coefficients and sets while coding, which increases its memory requirement as well as computational complexity. Also, it uses memory inefficient conventional discrete wavelet transform (DWT) to compute the transformed coefficients. In this study, a Low-Complexity Block Tree Coding (LCBTC) algorithm that uses two state-tables and two very small lists, is proposed. Similar to WBTC, it also uses sorting and refinement passes in each bit-plane. However, it encodes the coefficients in block-tree manner using depth-first search approach to reduce the computational complexity. It uses DWT coefficients obtained from modified fractional wavelet filter (MFrWF) rather than conventional DWT, which further reduces the overall memory and complexity of the image coder. The simulation results show that the memory requirement and computational complexity of LCBTC is much less than WBTC and other state-of-the-art coding algorithms. These features make the image coder a better candidate for compression in memory constrained and real-time visual sensor networks.

Brain tumour detection is still a challenging problem both in medical and image processing. In this study, two scenarios are applied to diagnose human brain tumours, classical image processing (CIP) and a new algorithm called binary image with variable fuzzy level (BIVFL). Magnetic resonance imaging (MRI) process from the BraTs 2019 database. Edge detection and prognosis area cropping are parts of the high-computational CIP algorithm. For the CIP, the best filter is the combination of the horizontal and vertical modes of the Prewitt filter. In the BIVFL, by changing the fuzzy level for binarisation images two clusters are filled with tumour and no tumour areas, and the tumour area is extracted in various accuracies. Sensitivity, specificity, precision, and accuracy of the BIVFL are varied according to the fuzzy level, and the best value of sensitivity is 0.9, but the value of three other parameters is 1 for an interval of fuzzy levels. The BIVFL is a simple and fast tumour area extraction algorithm, and also it is easy to implement with digital signal processors. Histogram and power signal-to-noise ratio of the BIVFL is remarkable.

Multiple Sclerosis (MS) causes the central nervous system to malfunction due to inflammation surrounding nerve cells. Detection of MS at an early stage is very important to prevent progressive MS attacks. Clinical findings, cerebrospinal fluid examinations, the evoked potentials, magnetic resonance imaging (MRI) findings have an important role in the diagnosis and follow-up of MS. However, many of the findings on MRI may indicate brain disorders other than MS. In addition, the clinical practices accepted by physicians for MS detection are very limited. In this study, a Mask R-CNN based method in two dataset is proposed for the automatic detection of MS lesions on magnetic resonance scans.We also improved the ROI detection stage with RPN in the Mask R-CNN to easily adapt for different lesion sizes. MS lesions in different sizes in the dataset are successfully detected with 84.90% Dice similarity rate and 87.03% precision rates using the proposed method. In addition, volumetric overlap error and lesion-wise true positive rate are obtained as 12.97% and 73.75%, respectively. Moreover, performance tests of the use of different numbers of GPU hardware structures are also performed and the evaluation of its effects on processing speed is performed on experimental studies..

Diagnostic medical imaging plays an imperative role in clinical assessment and treatment of medical abnormalities. The fusion of multimodal medical images merges complementary information present in the multi-source images and provides a better interpretation with improved diagnostic accuracy. This paper presents a CT–MR neurological image fusion method using an optimised biologically inspired neural network in nonsubsampled shearlet (NSST) domain. NSST decomposed coefficients are utilised to activate the optimised neural model using particle swarm optimisation method and to generate the firing maps. Low and high-frequency NSST subbands get fused using max-rule based on firing maps. In the optimisation process, a fitness function is evaluated based on spatial frequency and edge index of the resultant fused image. To analyse the fusion performance, extensive experiments are conducted on the different CT–MR neurological image dataset. Objective performance is evaluated based on different metrics to highlight the clarity, contrast, correlation, visual quality, complementary information, salient information, and edge information present in the fused images. Experimental results show that the proposed method is able to provide better-fused images and outperforms other existing methods in both visual and quantitative assessments.

Local feature extraction is one of the key characteristics of convolutional neural networks (CNNs). This study proposes an adaptive local feature enhancement (ALFE) model with a low-frequency general appearance-enhancement operator and a high-frequency local detail enhancement operator to improve local features of CNNs. Through supervised training, the model could adaptively adjust enhancement parameters and achieve a global-local enhancement of training images and CNNs. The performance of ALFE was first preliminarily evaluated with a self-built CNN on CIFAR-10 data set in different conditions of image augmentation and feature pooling. CNNs with ALFE could increase the top-1 accuracy compared against CNNs in the same conditions, and nearly reach the same level of performance for different pooling approaches. With only two extra adjustable parameters, this model could effectively avoid overfitting, without affecting the convergence speed of CNN. In addition, the extra burden of network complexity could be neglected. Further, experiments of three existing popular CNNs (AlexNet, VGGNet and ResNet) with ALFE were carried out on dogs versus cats, Tiny ImageNet, and SVHN data sets, respectively. The results show that ALFE is feasible for the existing popular CNN models, improving their top-1 accuracies without changing their convergence speeds.

This paper introduces a new HEp-2 cell classification model under two phases. In the initial phase, the input image is segmented using the morphological operations (opening and closing), and the segmented image is given for Convolutional Neural Network (CNN) classifier that gives the classified output. In the second phase, the given input is processed under (i) segmentation process (ii) Feature Extraction, and (iii) Classification. From the segmented images; the features like Gray level co-occurrence Matrix TGLCM) and Gray level Run Length Matrix (GLRM) are extracted. After extracting the features, they are subjected to a classification process, where Neural Network (NN) is used. Finally, the mean of both classified output (first phase and second phase) is considered to be the final classified output. As the main contribution, to enhance the classification accuracy, the hidden neurons of both classifiers (CNN and NN) are optimally chosen during the classification process. To make this possible, this paper aims to propose a new Randomized Update based Grey Wolf Optimization (RP-GWO) algorithm. Finally, the performance of the implemented approach is compared over other conventional approaches and its superiority is proven with respect to certain measures.

The performance of an image enhancer is usually evaluated either perceptually or functionally. The perceptual evaluation is carried out from a human perspective, i.e. by considering features related to human perception of image content and details. The functional evaluation is made instead from a machine perspective, i.e. by judging the enhancer effects with a specific machine application. his work proposes a comprehensive, empirical evaluation accounting for both perceptual and functional aspects. Precisely, 13 enhancers lowering undesired illumination effects are considered within the keypoint based image description and matching task, which is relevant to many computer vision fields. Each enhancer is first evaluated perceptually, then it is employed as a pre-processing step of the popular algorithms SIFT and ORB and judged by measuring how its use influences the performance of these algorithms. his study, conducted on a freely available data set, shows that the enhancement generally improves the perceptual features of the input image as well as the SIFT and ORB performance. More importantly, it reveals the existence of a correlation among some of their perceptual and functional measures. In this way, this work contributes to promote a more aware use of enhancement techniques within the mage description and matching task.

Accurately segmenting lungs from CT images is a fundamental step for quantitative analysis of lung diseases. However, it is still a challenging task because of some interferential factors, such as juxta-pleural nodules, pulmonary inflammation, as well as individual anatomical varieties. In this study, with the combination of a superpixel approach and a hybrid model composed of convolutional neural network and random forest (CNN-RF), the authors propose a novel algorithm to segment lungs from CT images in an automatic and accurate fashion. The authors' lung segmentation covers three main stages: image preprocessing, lung segmenting and segmentation refining. A lung CT image denoised with a fractional-order grey similarity approach is first segmented to a set of superpixels, and the CNN-RF model is then employed to classify the superpixels and identify lungs from the CT image. The segmentation result is further refined by separating the left and right lungs, eliminating trachea, and correcting lung contours. Experiments show that their algorithm can generate more accurate lung segmentation results with 94.98% Jaccard's index and 97.99% Dice similarity coefficient, compared with ground truths, and it achieved better results compared with several feature-based machine learning techniques and current methods on lung segmentation.

Under water images are likely to suffer from severe degradation such as colour distortion, low contrast, and fuzz content, caused by the absorption and scattering effects of the water. To improve the visual appearance of the image, the authors present an adaptive algorithm for effective underwater image enhancement using a randomly wired neural network (RWNN) and synergistic evolution (SE). In doing so, they sequentially conduct colours adjustment, contrast improvement and luminance enhancement while enhancing details by an edge-preserving technique. To set up the system, they develop a multi-strategy cooperating evolution algorithm to figure out the optimal parameter values. Extensive experimental results show that the proposed model improves both subjectively and quantitatively the quality of underwater images.

Due to the COVID-19 pandemic at present, it is necessary to detect whether pedestrians in public places wear face masks or not for preventing the spread of novel coronavirus. The pedestrian flow in public places is large, and it puts forward higher requirements for the accuracy and speed of real-time mask detection. Improving the face mask detection effect especially in the night environment is a challenging problem. A novel object detector namely MaskHunter is proposed in this study for the real-time mask detection. Specifically, the authors propose novel effective structures of backbone, neck and prediction head based on YOLOv4 series, which achieves the state-of-the-art performance and a novel improved Mosaic data augmentation method. Moreover, they propose a novel mask-guided module to enhance the discrimination ability of face mask especially in the night environment. As a consequent, experiments show that MaskHunter achieves better detection performance for real-time mask detection compared with other obtained models in this scenario.

In image watermarking, the quality of the watermarked image and the watermark's robustness are the main indicators for evaluating the performance of the watermarking scheme, but they contradict each other, making it difficult to design a rational watermarking method. In this study, the multi-objective evolutionary algorithm based on decomposition (MOEA/D) is employed to solve the multi-objective optimisation problem to ensure the balance between contradictory objectives. The watermark is embedded by selecting the regions with the lowest human visual characteristics in the host image and modifying the frequency coefficients of the discrete cosine transform. At this time, the determination of the positions of embedding coefficients and parameters related to the modifications results in the multi-objective optimisation problem on the robustness and imperceptibility, and the optimal solution is determined by finding the approximate Pareto set. Simulation results for various attacks on watermarked images demonstrate the effectiveness of the proposed watermarking technique.

The dual problem of image super-resolution (SR), which is referred to as compact-resolution (CR), and the corresponding image reconstruction are studied. These two problems have been studied independently by the researchers. In this study, a novel model for image CR and the corresponding reconstruction using the reversible network has been proposed. The reversible network has two properties, the first property, lossless information forwarding, which makes the compact-resolved image retain more information from the original HR image. The second property, bidirectional mapping, by which the forward and reverse propagation of a reversible network can be utilised to implement image CR and reconstruction, respectively, i.e. using the reverse process of image CR to guide the reconstruction. In addition, the utilisation of a reversible network may reduce the size of the model. The superiority of the proposed model was demonstrated by comparing its performance with the state-of-the-art methods on four well-known benchmark datasets.

Recent works show that the performance of convolutional neural networks (CNNs) could be improved by making the network wider and introducing residual connections. For example, the Inception architecture is one of the classical models. However, the structure of the series of Inception is complex, and there are more convolution layers and existing redundant image feature information. In order to further improve the performance and observability of CNNs, a novel type of network structure based on two modules, wider module and residual module, is proposed to make full advantage of the information of the image and learn more abundant features in this study, which is called residual-wider network (R-WN). The structure of R-WN is easy to understand and adopts modular design method. Experiential results demonstrate that the proposed R-WN with optimal structure and much fewer convolution layers could achieve much better image recognition results on four classical open data sets (MNIST, CIFAR10, SVHN, and Oxford Flowers 17).

To break the curtailment of digital imaging and retrieve appropriate information with different focused images, a novel edge-aware multi-focus image fusion is proposed by integrating the single-image super-resolution (SISR) method along with the edge-preserving filters. Initially, the multi-focus images are converted to high resolution images by estimating missing high frequency details from its blurred versions. With acquired high resolution images, smoothing by median and the anisotropic diffusion filters are performed to extract focused regions. An initial weight map is constructed by using maximum selection of pixel intensities of the difference images obtained with filtering. The precision of estimated weight map is further improved by exhibiting morphological operations and guided filter. Finally, the images are fused based on the optimised decision map. Simulation results of proposed fusion work are evaluated with seven metrics and the values are compared with different state-of-the-art methods. Both the quantitative and qualitative analyses showed the excellence of proposed work over other fusion methods.

Image regularised features play a critical role in image processing domain, by integrating regularised feature and structural similarity, a new full-reference image assessment method (IRF_SSIM) is proposed in this study. As well known, the gradient operator always be used to capture the edge information of the image, while the total variational regularised features can be adopted to calculate the detailed change information of image contrast and texture, as well as noise removal and edge retention. Therefore, the IRF_SSIM method extends the gradient features into the image regularised features to measure the structural changes in the image. In addition, image quality is also affected by variations of luminance and contrast. For a more comprehensive image quality assessment, the IRF_SSIM method considers the changes in structure, luminance and contrast simultaneously. In other words, the total image quality is estimated by structural similarity calculated by integrating the effects of image structure, luminance and contrast changes. Comparing with the representative methods, the experimental results illustrate that the IRF_SSIM method is highly consistent with the subjective assessment results.

An efficient fusion method of infrared and visible images is proposed based on the parallel saliency features. The method integrates significant feature information from the source images of multiple imaging modalities into a single fused image in a multi-scale domain, while suppressing visual artefacts and retaining more detail and texture information. First, the input images are decomposed into two-scale image representations, namely the base and detail layers, using a Gaussian filter. Second, the parallel saliency features of the high contrast and detail textures are captured to acquire the saliency maps. The contrast saliency weight map of the base layers based on the weighted local intensity energy aims to highlight the salient targets in infrared images and preserve the high-intensity regions in visible images, while the detail saliency weight map of the detail layers using the structure tensor to extract the detail texture information. Finally, the final image is reconstructed by the fused base and detail layers. Sufficient experimental results convincingly demonstrate that the presented method can achieve a comparable or superior performance compared with several state-of-the-art fusion methods via subjective assessments and objective evaluations, and it is more suitable for the practical applications due to the high computing efficiency.