Search

This paper presents a textile defect detection method that utilizes a multi‐proportion spatial attention mechanism and channel memory feature fusion network by addressing the difficulties presented by complicated shapes and large size variations. In particular, a multi‐proportion spatial attention mechanism (MPAM) is introduced, which employs multi‐proportion convolution to improve the backbone network's capacity to detect non‐uniform structural defects. Additionally, the generality and adaptability of the model are enhanced by a multi‐scale spatial pyramid pooling structure (MS‐SPP). Second, a channel attention mechanism‐based memory feature fusion network is developed, which incorporates channel attention to adaptive weight the feature channels, focusing on crucial information channels to efficiently fuse contextual features and enhance the model's memory capacity. Finally, a novel efficient Wise‐IoU (EWIoU) loss function is proposed, which utilizes a dynamic non‐monotonic focusing mechanism to increase the penalty on distance measurement, thus enhancing the model's detection performance. Experiment findings on the ZJU‐Leaper and Tianchi textile datasets reveal that compared to the YOLOv7 baseline, the method in this paper has an increase of 6.5 and 2 percentage points, respectively, and the detection accuracy is better than most existing networks. [ABSTRACT FROM AUTHOR]

Video‐based emotion recognition has been a long‐standing research topic for computer scientists and psychiatrists. In contrast to traditional discrete emotional models, emotion recognition based on continuous emotional models can better describe the progression of emotions. Quantitative analysis of emotions will have crucial impacts on promoting the development of intelligent products. The current solutions to continuous emotion recognition still have many issues. The original continuous emotion dataset contains incomplete data annotations, and the existing methods often ignore temporal information between frames. The following measures are taken in response to the above problems. Initially, aiming at the problem of incomplete video labels, the correlation between discrete and continuous video emotion labels is used to complete the dataset labels. This correlation is used to propose a mathematical model to fill the missing labels of the original dataset without adding data. Moreover, this paper proposes a continuous emotion recognition network based on an optimized temporal convolutional network, which adds a feature extraction submodule and a residual module to retain shallow features while improving the feature extraction ability. Finally, validation experiments on the Aff‐wild2 dataset achieved accuracies of 0.5159 and 0.65611 on the valence and arousal dimensions, respectively, by adopting the above measures. [ABSTRACT FROM AUTHOR]

Mapping high dynamic range (HDR) images to low dynamic range (LDR) display devices while performing enhancement processing is a challenging task. To address the issues of chromatic drift, the imbalance between preserving naturalness and enhancing details during the mapping process, this paper proposed a tone mapping operator (TMO) based on an improved guided filter with multi‐layered processing. Firstly, the operator tackled the problem of insufficient extraction of detail layer information in complex scene images by redefining the formula of the guided filter and introducing adaptive regularization parameters. This enhancement enabled more effective extraction of details compared to traditional fixed‐parameter guided filters. Secondly, the algorithm employed a centre‐surround approach to subdivide and processed the detail layers individually, effectively suppressing halos and artefacts. Thirdly, image entropy was utilized to guide the fusion and enhancement processing of the detail layers, preventing excessive or insufficient enhancement. Finally, this paper proposed an adaptive gamma correction algorithm and a chromatic consistency algorithm to improve overall contrast and hue saturation. The results demonstrate convincing performance in terms of subjective visual evaluation and image quality assessment. [ABSTRACT FROM AUTHOR]

License plate detection is an important task in Intelligent Transportation Systems (ITS) and has a wide range of applications in vehicle management, traffic control, and public safety. In order to improve the accuracy and speed of mobile recognition, an improved lightweight YOLOv5s model is proposed for license plate detection. First, an improved Stemblock network is used to replace the original Focus layer in the network, which ensures strong feature expression capability and reduces a large number of parameters to lower the computational complexity; then, an improved lightweight network, ShuffleNetv2, is used to replace the backbone network of the YOLOv5s, which makes the model lighter and ensures the detection accuracy at the same time. Then, a feature enhancement module is designed to reduce the information loss caused by the rearrangement of the backbone network channels, which facilitates the information interaction in the feature fusion process; finally, the low‐, medium‐ and high‐level features in the Shufflenetv2 network structure are fused to form the final high‐level output features. Experimental results on the CCPD dataset show that compared to other methods this paper obtains better performance and faster speed in the license plate detection task, in which the average precision mean value reaches 96.6%, and can achieve a detection speed of 43.86 frame/s, and the parameter volume is reduced to 5.07 M. [ABSTRACT FROM AUTHOR]

Pixel‐level labeling of regions of interest in an image is a key step in building a labeled training dataset for supervised deep learning networks of images. However, traditional manual labeling of cancerous regions in digital pathological images by doctors is time‐consuming and inefficient. To address this issue, this paper proposes an automatic labeling method for whole slide images, which combines clustering and edge detection techniques. The proposed method utilizes the multi‐level feature fusion model and the Long‐Short Term Memory network to discriminate the cancerous nature of the whole slide images, thereby improving the classification accuracy of the whole slide images. Subsequently, the automatic labeling of cancerous regions is achieved by integrating a density‐based clustering algorithm and an edge point extraction algorithm, both based on the discriminated results of the cancerous properties of whole slide images. The experimental results demonstrate the effectiveness of the proposed method, which offers an efficient and accurate solution to the challenging task of cancerous region labeling in digital pathological images. [ABSTRACT FROM AUTHOR]

This paper proposes a nonparametric point cloud filter to address the issue that existing point cloud filtering methods cannot retain important point cloud features after filtering and often require complex parameter adjustments. Firstly, a nonparametric clustering method is proposed to cluster various features of the point cloud and filter out isolated outliers. Then, the manifold distance truncation method is adopted to remove the outlier cluster generated by the point cloud clustering to complete point cloud filtering. Additionally, the proposed nonparametric clustering algorithm is compared with four of the latest clustering algorithms, including K‐means clustering and OPTICS clustering to verify the rationality of the clustering features. Finally, the filtering results of the nonparametric point cloud filter are compared with those of statistical filtering and two other recently proposed point cloud filters to demonstrate its improved filtering effect and algorithm stability. The experimental results indicate that the proposed nonparametric point cloud filter can achieve better‐filtering results and retain more point cloud features without adjusting parameters. [ABSTRACT FROM AUTHOR]

Tumour medical image segmentation plays a crucial role in clinical imaging diagnosis. Existing research has achieved good results, enabling the segmentation of three tumour regions in MRI brain tumour images. Existing models have limited focus on the brain tumour areas, and the long‐term dependency of features is weakened as the network depth increases, resulting in blurred edge segmentation of the targets. Additionally, considering the excellent segmentation performance of the Swin Transformer(Swin‐T) network, its network structure and parameters are relatively large. To address these limitations, this paper proposes a brain tumour segmentation framework with deep nuanced reasoning and Swin‐T. It is mainly composed of the backbone hybrid network (BHN) and the deep micro texture extraction module (DMTE). The BHN combines the Swin‐T stage with a new downsampling transition module called dual path feature reasoning (DPFR). The entire network framework is designed to extract global and local features from multi‐modal data, enabling it to capture and analyze deep texture features in multi‐modal images. It provides significant optimization over the Swin‐T network structure. Experimental results on the BraTS dataset demonstrate that the proposed method outperforms other state‐of‐the‐art models in terms of segmentation performance. The corresponding source codes are available at https://github.com/CurbUni/Brain‐Tumor‐Segmentation‐Framework‐with‐Deep‐Nuanced‐Reasoning‐and‐Swin‐T. [ABSTRACT FROM AUTHOR]

Combining edge devices with intelligent inspection for transmission lines can fulfill the demand for real‐time defect detection in the field. However, there has been limited research on algorithms suitable for edge devices with low computational power and memory, and the existing research primarily focuses on CPU optimization. To address these issues, this paper proposes a real‐time defect detection method for transmission line endpoints based on YOLO‐GSS (YOLOv8 with Mosaic‐9, G‐GhostNet, S‐FPN, and Spatial Intersection over Union (SIoU) modifications). First, the authors improve the input of the YOLOv8 network using Mosaic‐9 to increase the number of input features in the training phase and enhance algorithm robustness. Next, the authors introduce G‐GhostNet and S‐FPN to enhance the backbone and neck sections while improving inference speed and accuracy. Finally, the authors modify the Complete Intersection over Union loss function of YOLOv8 using SIoU to further improve the detection accuracy. Experimental results demonstrate that compared to the original YOLOv8, the proposed method achieves a 5x increase in inference speed on Nvidia Jetson NX edge devices and a 7.7% improvement in accuracy, meeting the real‐time defect detection requirements for transmission line field inspections. [ABSTRACT FROM AUTHOR]

Video steganography has the potential to be used to convey illegal information, and video steganalysis is a vital tool to detect the presence of this illicit act. Currently, all the motion vector (MV)‐based video steganalysis algorithms extract feature sets directly from the MVs, but ignoring the embedding operation may perturb the statistical distribution of other video encoding elements, such as the skipped macroblocks (no direct MVs). This paper proposes a novel 11‐dimensional feature set to detect MV‐based video steganography based on the above observation. The proposed feature is extracted based on the skipped macroblocks by recompression calibration. Specifically, the feature consists of two components. The first is the probability distribution of motion vector prediction (MVP) difference, and the second is the probability distribution of partition state transfer. Extensive experiments on different conditions demonstrate that the proposed feature set achieves good detection accuracy, especially in lower embedding capacities. In addition, the loss of detection performance caused by recompression calibration using mismatched quantization parameters (QP) is within the acceptable range, so the proposed method can be used in practical scenarios. [ABSTRACT FROM AUTHOR]

This paper focuses on the intelligent detection of prohibited items in X‐ray images during the security checking process. An intelligent semantic segmentation model of prohibited items in X‐ray images is proposed based on the attention‐based object localization method. Based on the pre‐trained CNN classification framework, the attention mechanism can map the high‐layer semantic information of objects into the input space, while generating energy saliency maps to locate the prohibited items. In order to make the obtained attention maps discriminative, the lateral and contrastive inhibition strategies are introduced and combined together which can highlight the responses of activated neurons. Under the guidance of attention responses, two traditional image segmentation algorithms are employed to achieve the semantic segmentation results for the prohibited items detection in X‐ray images. The proposed semantic segmentation model relies on weakly supervised learning mechanism, and only depends on the category labels of prohibited items, which greatly avoids the work cost of data semantic annotation. The experimental results based on the public SIXray baseline and the self‐built X‐ray image database demonstrate the proposed method can achieve about 65% IoU localization precise averagely. In addition, comparison experiments were carried out with the state‐of‐the‐arts and ablation experiments to verify the effectiveness of the proposed model. [ABSTRACT FROM AUTHOR]

The YOLO series of algorithms have made substantial contributions to the detection of insulator defects in power transmission line operations. However, existing target detection algorithms for the small target detection and low‐quality insulator images encounter difficulties in effectively capturing relevant features, resulting in a higher probability of target loss. To identify and classify defects in the operational state of insulators, an improved YOLOv8 target identification algorithm called DGW‐YOLOv8 is proposed in this paper. The deformable attention backbone of the DGW‐YOLOv8 target identification algorithm is designed by adding the deformable ConvNets v2 module and the global attention mechanism. This addition reduces the feature loss caused by the network feature processing, enhances the sensitivity of the algorithm to small‐scale targets, and reduces the impact caused by the different global positions of the targets. Additionally, to address the problem of low quality of captured images, WIoU v3 is used to replace CIoU in the original YOLOv8 target identification algorithm to optimize the loss function, reduce the degrees of freedom, and improve the network robustness. Experimental results demonstrate that the enhanced YOLOv8 algorithm can achieve an improvement of 2.4% and 5.5% in mAP and mAP50‐95, respectively, compared with the original algorithm. [ABSTRACT FROM AUTHOR]

The images captured by forest nighttime imaging devices are prone to degradation and noise issues due to environmental constraints. This paper proposes a novel framework based on the self‐calibrating illumination framework to improve image quality. The framework consists of four specific steps: (1) addressing non‐uniform degradation by estimating illuminance components through non‐uniform MaxRGB, (2) extracting rich information using multi‐scale inflationary convolution, (3) analyzing the relationship between different input levels and the initial level and progressively adjusting the luminance components using a channel self‐calibration module to achieve continuous improvement in luminance, and (4) designing a reflectance optimization module that enhances illuminance while suppressing noise through a regularization model. To evaluate the proposed method, a dataset comprising 1000 forest night images is constructed using an image simulation strategy. The effectiveness of this method was validated on both synthetic and real datasets, demonstrating its ability to significantly enhance nighttime forest images. When compared to six other algorithms, this method outperformed them in eight evaluation metrics, including a peak signal‐to‐noise ratio metric of 22.839 and a structural similarity metric of 0.909. The experimental results clearly indicate that this method excels in enhancing image quality and fidelity. [ABSTRACT FROM AUTHOR]

Image inpainting aims to repair the damaged region according to the known content in the damaged image. Recently, image inpainting methods have poor effects on high‐resolution damaged images, and the research on the inpainting of large‐area damaged images is limited. Therefore, this paper proposes an image inpainting method based on Generative Adversarial Networks (GAN) inversion and autoencoder. This work consists of two phases: first, the authors design an autoencoder‐based GAN, which learns the mapping from noise to low‐dimensional feature maps by training a generator, and then converts the generated feature maps into high‐resolution images. Thus, the difficulty of learning the mapping relationship is reduced. Second, the authors adopt the learning‐based GAN inversion to infer the closest latent code. The trained GAN is then used to reconstruct the complete image. Finally, the authors compare their method with other classical methods on the CelebAMask‐HQ, Flickr‐Faces‐HQ, and ImageNet datasets. According to the quantitative comparison, when the mask range is large, in other words, when the image has a large area of damage, the authors' method is superior to the comparison methods. According to the qualitative comparison, the structure of the high‐resolution image inpainted by the authors' method is more reasonable and the texture details are more realistic. [ABSTRACT FROM AUTHOR]

In real scenarios, accurate and real‐time detection of traffic signs is of great significance to the automatic driving system. To meet the requirements of detection accuracy and speed, a new traffic sign detection method based on YOLOv5 and Swin‐Transformer is proposed in this paper. First, based on the traditional Focus structure, a lightweight shallow information enhancement module is designed. Second, to enhance the channel weights of useful information, an adjustable channel attention mechanism is proposed. Additionally, a Cross Stage Partial module based on Swin‐Transformer is designed to capture contextual information around traffic signs, thereby improving the detection accuracy of small‐scale traffic signs. Finally, to better fuse deep semantic features and shallow detail features, an adaptive feature fusion method is proposed. To verify the superiority of the proposed method, experimental verification was carried out on TT100K and DFG traffic sign detection datasets, and their mAP, AP50 and FPS were (75.3, 94.8, 82) and (79.7, 85.9, 118), respectively. The experimental results show that the proposed traffic sign detection method has high accuracy and real‐time performance, and can meet the needs of traffic sign detection in the actual scene. [ABSTRACT FROM AUTHOR]

In this paper, an additive margin network with adaptive feature recalibration, AMNets, for handling a wide range of visual tasks, is proposed. The new AMNets consists of three improvements on convolutional neural networks: (1) squeeze‐and‐excitation mechanism was embedded into the the residual block of single layer to improve the performance of the classification, (2) softmax loss was replaced by the additive margin softmax loss to minimize intra‐class variance and maximize inter‐class variance, and (3) Adam optimizer was exploited to update iteratively the network parameters. Experiments on the Brain Stroke CT Image Dataset show that our additive margin network is quite effective to improve state‐of‐the‐art algorithms. Network algorithms have been extensively used in medical image field to pursuit a more accuracy of disease recognition. This paper proposes AMNets, an additive margin network with adaptive feature recalibration for handling a wide range of visual tasks in medicine. [ABSTRACT FROM AUTHOR]

The bridge bottom crack detection provides important state information for bridge disease control and safety assessment. This paper proposes a detection method based on deep learning Faster R‐CNN and BIM (Building Information Modeling). The UAV (Unmanned Aerial Vehicle) was used for close aerial photography to obtain high‐resolution crack images of the concrete surface at the bottom of a bridge. Through deep learning algorithms, a Faster R‐CNN model is trained and established for crack identifications. The crack identification accuracy rate and recall rate reach 92.03% and 96.54%, respectively. Crack images are mapped to a BIM model developed for the chosen bridge, and the box girder family with cracks and the three crack families of transverse cracks, longitudinal cracks and turtle cracks are established. The cracks are located and the visualization of the beam bridge with cracks was completed. In order to better assess the health condition of the bridge. The results show that the combination of UAV bridge crack detection and modelling solves the remote, visual and automated detection of cracks on the surface of bridge structures, which are difficult to reach manually, and has important scientific research and engineering application value. [ABSTRACT FROM AUTHOR]

Recent years, the development of DeepFake has raise a lot of security problems. Therefore, detection of DeepFake is critical. However, the existing DeepFake detection methods are often vulnerable to adversarial attacks, i.e. adding carefully crafted imperceptible perturbations into forged images is possible to evade detection. In this paper, a DeepFake detection method based on image denoising is proposed by combining variational autoencoder (VAE) and generative adversarial network (GAN), namely D‐VAEGAN. Firstly, an encoder is designed to extract the features of the image in a low‐dimensional latent space. Then, a decoder reconstructs the original clean image using the features in this low‐dimensional latent space. Secondly, an auxiliary discriminative network is introduced to further improve the performance of the model, which improves the quality of the reconstructed images. Furthermore, feature similarity loss is added as a penalty term to the reconstruction optimization function to improve the adversarial robustness. Experimental results on the FaceForensics++ dataset show that the proposed approach significantly outperforms the five adversarial training‐based defence methods. The approach achieves 96% in accuracy, which is on average about 50% higher than other comparison methods. [ABSTRACT FROM AUTHOR]

A novel image encryption algorithm based on hyperchaos and DNA encoding is proposed in this paper. Unlike the traditional DNA‐based encoding method, a more fine‐grained encryption approach is adopted, which enhances the security of the system. To address the inherent binary arithmetic limitation of traditional DNA computing, the authors develop a novel symmetric and reversible DNA subtraction operation that fully utilizes the encryption advantage of DNA computing. To enhance security, the authors introduce a DNA crossover mechanism based on pixel, which can generate different ciphertexts even for the same pixel values. In the algorithm, the authors propose a diffusion strategy that combines left diffusion, right diffusion, and XOR parameters. This strategy not only ensures the avalanche effect, where even a single pixel change in the plaintext image will cause a drastic change in the ciphertext image, but also greatly enhances the security of the algorithm. Experimental results and performance analysis indicate that the authors' proposed system has good encryption performance and the ability to resist noise attacks and clipping attacks. [ABSTRACT FROM AUTHOR]

The network structure of digital watermarking algorithm based on deep learning is usually encoder‐noise layer‐decoder. Most of the existing encoders suffer from the problem of insufficient feature extraction, and the introduction of simulated differentiable joint photographic experts group (JPEG) compression in the noise layer cannot ensure the robustness under real JPEG. In this paper, a watermarking algorithm based on multi‐scale auto‐encoder is proposed, which can effectively extract the image feature information by combining with the channel attention mechanism. At the same time, some parameters of decoder and encoder are shared to reduce redundant feature embedding and improve extraction accuracy. This paper also proposes a robust training scheme against JPEG compression, which can guide the model to store the watermark in the low‐frequency region needed for decoding. Experimental results show that the peak signal‐to‐noise ratio (PSNR) of the proposed algorithm is above 48 and the decoding rate is above 99% under JPEG compression with quality factor Q = 50. Moreover, this scheme can effectively promote the combination of noise layer in training. In addition, the proposed algorithm is also robust to other common network noises. [ABSTRACT FROM AUTHOR]

To develop a fast and effective image deblurring method, the blind recovery of motion‐blurred images based on DeblurGAN(GAN, Generative Adversarial Networks) is researched.Firstly, the number of residual modules in the DeblurGAN network is changed, and an attempt is made to optimize the network structure to achieve better results in the blind recovery of motion‐blurred images. Secondly, an image deblurring method based on the improved DeblurGAN network is proposed.This paper makes corresponding adjustments to the generator and discriminator networks to change their input and output size to 512 × 512 while keeping the overall structure of the network unchanged, and the experimental results show that the quality of the restored images has been greatly improved. In addition, the images were divided into four classes of images,to achieve improved restoration of blurred images with specific content. Experimental results on the benchmark GoPro dataset validate that the improved DeblurGAN achieves more than 1.5 dB improvement than DeblurGAN in terms of peak signal‐to‐noise ratio (PSNR) as trained utilizing the same amount of data, and structural similarity assessment (SSIM) evaluation means and maximum values increased between 0.2 and 0.3. The improved DeblurGAN is more effective in terms of both blur removal and detail recovery. [ABSTRACT FROM AUTHOR]

Most existing domain adaption methods solve the problem of data distribution similarity in single‐source to single‐target domain or multi‐source to single‐target domain adaption, but the more realistic multi‐source to multi‐target domain transfer scenarios are ignored. At the same time, less attention is paid to the similarity of task distribution among domains and the corresponding relationship between coarse‐grained and fine‐grained, which leads to the problem of low classification accuracy. In this paper, a multi‐source to multi‐target domain adaption (MSMTDA) method based on similarity measurement is proposed. Firstly, a cross‐domain similarity measurement mechanism based on the idea of data distribution similarity and task distribution similarity is set up to judge and optimize the distribution difference among source domains and target domains. Secondly, global‐local‐bidirectional alignment is used to solve the problem of poor transfer performance in the existing distributed alignment. Then, the cross‐domain mutual learning strategy is proposed, which reduces classifier discrepancy caused by decision boundaries and distribution difference in multiple domains. Finally, the proposed method is verified on three benchmark datasets in image classification. The experimental results show that the proposed method performs better in both classification accuracy and training time when compared with the conventional method. [ABSTRACT FROM AUTHOR]

Whole brain segmentation from magnetic resonance images (MRI) is crucial in diagnosing brain diseases and analyzing neuroimaging data. Despite advances through deep learning, challenges such as uneven gray distribution and the presence of artifacts still present hurdles in medical image processing. These limitations are often a result of insufficient spatial contextual information and lack of attention to important regions within existing models. To address these issues, this paper presents SSA‐UNet (Squeeze‐and‐Excitation and Self‐Attention UNet), a uniquely designed deep convolutional neural network that integrates spatial constraints by converting three consecutive 2D MRI slices into a single 2.5D image. This facilitates capturing inter‐slice dependencies effectively. Additionally, the newly formulated SSA block, which sequentially incorporates channel attention and Self‐Attention mechanisms, is placed before the decoders in the conventional U‐Net architecture. This enables the network to automatically weight different feature maps and focus more effectively on regions requiring precise segmentation. Rigorous evaluations on LPBA40 and IBSR18 datasets substantiate the remarkable improvements in accuracy and stability achieved by SSA‐UNet. Results indicate Dice coefficients of 98.38% and 97.47%, specificity of 99.69% and 99.57%, and sensitivity of 98.5% and 97.98% for the respective datasets. Compared to other existing models, SSA‐UNet shows significant improvements on both the LPBA40 and IBSR18 datasets. On the LPBA40 dataset, SSA‐UNet's Dice coefficient improved by 0.33% compared to the sub‐optimal model, while on the IBSR18 dataset, the improvement reached 1.78%. These empirical findings demonstrate SSA‐UNet's heightened capability in addressing the long‐standing challenges in MRI‐based whole‐brain segmentation. [ABSTRACT FROM AUTHOR]

In the field of scientific visualization, the upscaling of time‐varying volume is meaningful. It can be used in in situ visualization to help scientists overcome the limitations of I/O speed and storage capacity when analysing and visualizing large‐scale, time‐varying simulation data. This paper proposes self‐attention residual network‐based spatial super‐resolution (SARN‐SSR), a spatial super‐resolution model based on self‐attention residual networks that can generate time‐varying data with temporal coherence. SARN‐SSR consists of two components: a generator and a discriminator. The generator takes the low‐resolution volume sequences as the input and gives the corresponding high‐resolution volume sequences as the output. The discriminator takes both synthesized and real high‐resolution volume sequence as the input and gives a matrix to predict the realness as the output. To verify the validity of SARN‐SSR, four sets of time‐varying volume datasets are applied from scientific simulation. In addition, SARN‐SSR is compared on these datasets, both qualitatively and quantitatively, with two deep learning‐based techniques and one traditional technique. The experimental results show that by using this method, the closest time‐varying data to the ground truth can be obtained. [ABSTRACT FROM AUTHOR]

Most existing multi‐view stereo (MVS) methods fail to consider global context information in the stage of feature extraction and cost aggregation. As transformers have shown remarkable performance on various vision tasks due to their ability to perceive global contextual information, this paper proposes a transformer‐based feature enhancement network (TF‐MVSNet) to facilitate feature representation learning by combining local features (both 2D and 3D) with long‐range contextual information. To reduce memory consumption of feature matching, the cross‐attention mechanism is leveraged to efficiently construct 3D cost volumes under the epipolar constraint. Additionally, a colour‐guided network is designed to refine depth maps at a coarse stage, hence reducing incorrect depth predictions at a fine stage. Extensive experiments were performed on the DTU dataset and Tanks and Temples (T&T) benchmark and results are reported. [ABSTRACT FROM AUTHOR]

Existing segmentation‐based methods have made considerable progress in arbitrarily shaped text detection due to the advantage of dealing with shape variation. However, there still exist challenges to detecting accurate text instances with dense layouts, inaccurate annotations, and complex backgrounds. Many recent works have focused on improving arbitrary boundary prediction, but it may be difficult to accurately distinguish each instance of dense layouts because their boundary pixels may be mistakenly classified to produce inaccurate results (i.e., adhesive texts) with inaccurate annotation and complex backgrounds. Considering the local and long‐range dependencies, this paper proposes an efficient text detector, namely T‐Skeleton, to obtain more reliable segmentation detections. In the spirit of object skeletonization, we introduce the text instance skeleton highlighting the semantically significant structure (similar to the skeleton of a fish) to explicitly capture the long‐range dependencies of text instances. The key idea of T‐Skeleton is to calibrate the coarse text proposals by embedding text instance skeletons to separate crowd texts accurately and robustly. We further design a channel attention module to enlarge the performance margin between T‐Skeleton and the segmentation baseline. Experimental results on four publicly available datasets show the superiority of T‐Skeleton in handling long and curved texts. [ABSTRACT FROM AUTHOR]

Scene text in natural images carries additional semantic information to aid in image classification. Existing methods lack full consideration of the deep understanding of the text and the visual text relationship, which results in the difficult to judge the semantic accuracy and the relevance of the visual text. This paper proposes image classification based on Cross modal Knowledge Learning of Scene Text (CKLST) method. CKLST consists of three stages: cross‐modal scene text recognition, text semantic enhancement, and visual‐text feature alignment. In the first stage, multi‐attention is used to extract features layer by layer, and a self‐mask‐based iterative correction strategy is utilized to improve the scene text recognition accuracy. In the second stage, knowledge features are extracted using external knowledge and are fused with text features to enhance text semantic information. In the third stage, CKLST realizes visual‐text feature alignment across attention mechanisms with a similarity matrix, thus the correlation between images and text can be captured to improve the accuracy of the image classification tasks. On Con‐Text dataset, Crowd Activity dataset, Drink Bottle dataset, and Synth Text dataset, CKLST can perform significantly better than other baselines on fine‐grained image classification, with improvements of 3.54%, 5.37%, 3.28%, and 2.81% over the best baseline in mAP, respectively. [ABSTRACT FROM AUTHOR]

This paper proposes a super‐resolution reconstruction model, SRPGANto improve the visual quality of images based on generative adversarial networks (GANs) by improving the network structures of the generator and the discriminator. In the generator, a dual branch residual block is designed instead of the residual block, including a branch with an attention mechanism and a branch without an attention mechanism, to extract more differentiated features. Normalization methods are explored to avoid unstable training and bath normalization artifacts and use a half instance normalization layer that is more suitable for underlying visual problems compared with traditional batch normalization. In the discriminator, PatchGAN is applied instead of typical GAN to improve the generation of local texture by discriminating each patch rather than the global image. The experimental results on the public datasets demonstrate that the proposed SRPGAN can achieve excellent quantitative evaluation while improving the visual quality of reconstructed images. [ABSTRACT FROM AUTHOR]

While recent years have witnessed the unprecedented success of deep convolutional neural networks (CNNs) and vision transformers in single‐image super‐resolution (SISR), the degradation assumptions are simple and usually bicubic downsampling. Thus, their performances will drop dramatically when the actual degradation does not match this assumption, and they lack the capability to handle multiple degradations (e.g. Gaussian noise, bicubic downsizing, and salt & pepper noise). To address the issues, in this paper, the authors propose a joint SR model (JIRSR) that can effectively handle multiple degradations in a single model. Specifically, the authors build the parallel Transformer and CNN branches that complement each other through bidirectional feature fusion. Moreover, the authors also adopt a random permutation of different kinds of noise and resizing operations to build the training datasets. Extensive experiments on classical SR, denoising, and multiple degradation removal demonstrate that the authors' JIRSR achieves state‐of‐the‐art (SOTA) performance on public benchmarks. Concretely, the authors' JIRSR outperforms the second‐best model by 0.23 to 0.74 dB for multiple degradations removal and is 0.20 to 0.36 dB higher than the SOTA methods on the Urban100 dataset under the ×4 SR task. [ABSTRACT FROM AUTHOR]

In recent years, the task of maritime emergency rescue has increased, while the cost of time for traditional methods of search and rescue is pretty long with poor effect subject to the constraints of the complex circumstances around the sea, the effective conditions, and the support capability. This paper applies deep learning and proposes a YOLOv5s‐SwinDS algorithm for target detection in distress at sea. Firstly, the backbone network of the YOLOv5s algorithm is replaced by swin transformer, and a multi‐level feature fusion module is introduced to enhance the feature expression ability for maritime targets. Secondly, deformable convolutional networks v2 (DCNv2) is used instead of traditional convolution to improve the recognition capability for irregular targets when the neck network features are output. Finally, the CIoU loss function is replaced with SIoU to reduce the redundant box effectively while accelerating the convergence and regression of the predicted box. Experimenting on the publicly dataset SeaDronesSee, the Precision$Precision$, Recall$Recall$, mAP0.5$mAP_{0.5}$ and mAP0.5−0.95$mAP_{0.5-0.95}$ of YOLOv5s‐SwinDS model are 87.9%, 75.8%, 79.1% and 42.9%, respectively, which get higher results than the original YOLOv5s model, the YOLOv7 series of models, and the YOLOv8 series of models. The experiments verifies that the algorithm has good performance in detecting maritime distress targets. [ABSTRACT FROM AUTHOR]

Deep convolutional neural networks based remote sensing change detection has recently shown significant performance improvement. However, small region changes and global‐local features in high‐resolution remote sensing images are not fully explored. This paper introduces a hybrid U‐shaped and transformer network for change detection in high‐resolution remote sensing images. Specifically, a UNet++‐based backbone to facilitate feature learning across different scales. In addition, we introduce a transformer‐based feature fusion module for extracting long‐range dependencies, which can enhance the representation ability of the network. Furthermore, the introduced efficient channel attention mechanism can efficiently calibrate the feature representation and concentrate on more important feature information. Thanks to the above designs, the proposed method enjoys a strong ability to extract local and global features for remote sensing change detection. Extensive experimental results on different remote sensing images show that our method can achieve superior performance in comparison with state‐of‐the‐art change detection methods. [ABSTRACT FROM AUTHOR]

With the growing significance of data privacy protection, Source‐Free Domain Adaptation (SFDA) has gained attention as a research topic that aims to transfer knowledge from a labeled source domain to an unlabeled target domain without accessing source data. However, the absence of source data often leads to model collapse or restricts the performance improvements of SFDA methods, as there is insufficient true‐labeled knowledge for each category. To tackle this, Source‐Free Active Domain Adaptation (SFADA) has emerged as a new task that aims to improve SFDA by selecting a small set of informative target samples labeled by experts. Nevertheless, existing SFADA methods impose a significant burden on human labelers, requiring them to continuously label a substantial number of samples throughout the training period. In this paper, a novel approach is proposed to alleviate the labeling burden in SFADA by only necessitating the labeling of an extremely small number of samples on a one‐time basis. Moreover, considering the inherent sparsity of these selected samples in the target domain, a Self‐adaptive Clustering‐based Active Learning (SCAL) method is proposed that propagates the labels of selected samples to other datapoints within the same cluster. To further enhance the accuracy of SCAL, a self‐adaptive scale search method is devised that automatically determines the optimal clustering scale, using the entropy of the entire target dataset as a guiding criterion. The experimental evaluation presents compelling evidence of our method's supremacy. Specifically, it outstrips previous SFDA methods, delivering state‐of‐the‐art (SOTA) results on standard benchmarks. Remarkably, it accomplishes this with less than 0.5% annotation cost, in stark contrast to the approximate 5% required by earlier techniques. The approach thus not only sets new performance benchmarks but also offers a markedly more practical and cost‐effective solution for SFADA, making it an attractive choice for real‐world applications where labeling resources are limited. [ABSTRACT FROM AUTHOR]

This paper proposes a new method for hierarchical image segmentation based on the nonsymetry and anti‐packing pattern representation model (NAM) and the hierarchical density‐based spatial clustering of application with noise (HDBSCAN). The proposed framework consists of two phases. In the first phase, a super‐pixel generation algorithm base on NAM is proposed. In the second phase, instead of defining an affinity matrix to merge similar regions using spatial clustering, the distance matrix defined by different region features is directly fitted into an HDBSCAN clustering module in order to merge similar regions efficiently. Similar adjacent regions can be merged into larger ones progressively and form a segmentation dendrogram for image segmentation with the clustering module. The experiments show that the proposed algorithm has a comparable or even better performance compared to the state‐of‐the‐art hierarchical image segmentation algorithms while having much less time and memory consumption. [ABSTRACT FROM AUTHOR]

Computer vision‐based fire detection algorithms have been extensively researched over the past decades as they offer reliable and efficient solutions to early fire detection, as well as economical solutions in outdoor environments. The proposed systems have different levels of complexity; nevertheless, they rely heavily on colour based fire segmentation as the first processing stage using different colour spaces. This paper presents a new flame‐based colour space which accentuates the difference between fire and non‐fire pixel values. The hybrid Artificial Bee Colony‐Teaching Learning‐based Optimisation (ABC‐TLBO) and the k‐medoids clustering algorithm along with a Feature image constructed manually from a wide dataset of fire and non‐fire pixels were used to deduce the optimised colour space conversion matrix. The latter can then be used to transform the RGB pixels from a camera feed into the new Flame‐based colour space for a more efficient fire segmentation. The proposed colour space is capable to separate fire and non‐fire pixels into two intensity classes, whereby the resulting between‐class variance is significantly widened; this maximises the effectiveness of the binary thresholding to separate fire from non‐fire pixels. The experimental results demonstrate that the proposed flame‐based colour space provides a superior average qualitative and quantitative performance on benchmark datasets compared to state‐of‐the‐art colour models; a 20% average increase in F1 score demonstrates a better balance between fire detection and false alarm rates. Moreover, the computation time of the fire detection system is faster by 15.6% in average. The proposed system provides a better implementation of the colour‐based fire segmentation processing stage in vision‐based fire detection systems; it vindicates, in this application context, the superiority of flame‐tuned colour space over conventional ones. [ABSTRACT FROM AUTHOR]

Aircraft coating process has been an important part in manufacturing process of modern aviation products. For coating defect detection, the manual observation with naked eyes is usually utilized, which leads to low production efficiency. In this paper, the authors propose the improved YOLOv5‐OBB with the channel‐spatial attention block (CSAB), feature pyramid non‐local module (FPNM) and structured sparsity slimming criterion (SSSC). The CSAB can pay more attention to effective channel information features from the channel dimension and the target information area from the spatial dimension. The effective non‐local module called FPNM is proposed to further improve the detection accuracy. The authors utilize the oriented bounding boxes (OBB) to reduce redundant background information for coating defect detection. In addition, the SSSC is proposed to achieve network slimming and trade‐off between the efficiency and accuracy. The experimental results on several datasets demonstrate the effectiveness of the authors' scheme, which achieves superior performance. [ABSTRACT FROM AUTHOR]

The low‐light environment is integral to everyday activities but poses significant challenges in object detection. Due to the low brightness, noise, and insufficient illumination of the acquired image, the model's object detection performance is reduced. Opposing recent studies mainly developing using supervised learning models, this paper suggests LIDA‐YOLO, an approach for unsupervised adaptation of low‐illumination object detectors. The model improves the YOLOv3 by using normal illumination images as the source domain and low‐illumination images as the target domain and achieves object detection in low‐illumination images through an unsupervised learning strategy. Specifically, a multi‐scale local feature alignment and global feature alignment module are proposed to align the overall attributes of the image and feature biases such as background, scene, and target layout are thus reduced. The experimental results of LIDA‐YOLO on the ExDark dataset achieved the highest performance mAP score of 56.65% compared to several current state‐of‐the‐art unsupervised domain adaptation object detection methods. Compared to I3Net, the performance improvement is 4.04%, and compared to OSHOT, the performance improvement is 6.5%. LIDA‐YOLO achieves a performance improvement of 2.7% compared to the supervised baseline method YOLOv3. Overall, the suggested LIDA‐YOLO model requires fewer samples and presents a stronger generalization ability than previous works. [ABSTRACT FROM AUTHOR]

Tuberculosis (TB) is an extremely contagious illness caused by Mycobacterium tuberculosis. Chest tuberculosis classification is conducted based on a deep convolutional neural network architecture. In this research, a pre‐trained network is utilized to demonstrate the advantage of using the oversampling technique on the classification of TB and compare results with recent research that used the same dataset. Therefore, the dataset consists of 3500 uninfected TB cases and 700 infected with TB. This paper circumvents the imbalance by using the oversampling technique in X‐ray TB images to be fed into several pre‐trained networks for TB classification. The oversampling technique is crucial in enhancing the performance of TB classification compared with other pre‐trained models reported here. Inceptionv3 shows a promising result compared to other pre‐trained models; it achieves 99.94% accuracy, 99.88% precision, 100% recall, and 99.94% F1‐Score. [ABSTRACT FROM AUTHOR]

Visual surveillance systems have been playing an important role in monitoring and managing at public areas. However, the computational complexity of video compression in these applications is still a great challenge. To meet practical requirements, the authors propose in this paper a low‐complexity surveillance video coding solution in which the most recent Versatile Video Coding (VVC) standard is improved with a novel learning adaptive motion search algorithm. The proposed algorithm is designed based on the temporal motion and spatial texture characteristics of surveillance videos. First, the authors study and define a list of spatial and temporal features which indicates the motion and texture characteristics of surveillance video. These features are used together with a machine learning algorithm to appropriately assign a search range for the VVC motion search. Second, to reduce search points, the authors propose an adaptive Test Zone (TZ) search in which TZ steps are early terminated following the variation of spatial–temporal features. Performance evaluation conducted for a rich set of surveillance videos and relevant benchmarks have shown the superiority of the proposed method, notably with around 33% of encoding time saving when compared with the state‐of‐the art VVC solution and relevant benchmarks while asking for negligible compression loss. [ABSTRACT FROM AUTHOR]

Inverse halftoning is a technology that converts a binary image into a continuous tone image. Due to the wide application of inverse halftoning, many scholars have proposed several deep convolutional neural networks (DCNN) to optimize their performance. According to the observation, there is still room for improvement in content generation and detail recovery of the inverse halftone images generated by using the existing methods. Therefore, an inverse halftoning method based on supervised DCNN is proposed in this paper. The method consists of two parts: the multi‐level feature extraction model uses the down‐sampling to extract the features from the halftone image and remove the halftone noise dots on flat areas, which is implemented by four convolutional layers; the image reconstruction model uses up‐sampling to reconstruct image information, which is realized by four convolutional layers and two dense residual blocks. At the same time, in order to further recover the details, the down‐sampling feature maps and up‐sampling feature maps of the same size are concatenated by addition layers. Experimental results show that compared with other methods, the inverse halftone images obtained by the proposed network have better results in both subjective and objective evaluations. [ABSTRACT FROM AUTHOR]

Hazy images suffer from two problems. The low contrast can be enhanced by estimating a transmission layer, and the colour cast can be restored by estimating an airlight. These two variables, together with the albedo layer, are the constitutive elements of a hazy image. The resulting quality of dehazed images is inextricably linked to the accurate estimation of these components. However, it is ill‐posed to decompose these variables from a single image. As such, this paper presents an innovative algorithm intended to facilitate the optimal decomposition of a hazy image. Using the Markov random field model, an optimal framework is established that allows the simultaneous estimation of the three components across the three‐colour channels. To improve the visual quality, three improvements are proposed in the variational solution for the optimal components. The dehazed result is recomposed from the components with the transmission enhanced to circumvent any potential artefacts or information loss. Extensive experiments on natural images corroborate that the proposed algorithm outperforms state‐of‐the‐art dehazing methods, both qualitatively and quantitatively. [ABSTRACT FROM AUTHOR]

Visible‐infrared person re‐identification (VI‐ReID) aims to recognize images of the same person captured in different modalities. Existing methods mainly focus on learning single‐granularity representations, which have limited discriminability and weak robustness. This paper proposes a novel dual‐granularity feature fusion network for VI‐ReID. Specifically, a dual‐branch module that extracts global and local features and then fuses them to enhance the representative ability is adopted. Furthermore, an identity‐aware modal discrepancy loss that promotes modality alignment by reducing the gap between features from visible and infrared modalities is proposed. Finally, considering the influence of non‐discriminative information in the modal‐shared features of RGB‐IR, a greyscale conversion is introduced to extract modality‐irrelevant discriminative features better. Extensive experiments on the SYSU‐MM01 and RegDB datasets demonstrate the effectiveness of the framework and superiority over state‐of‐the‐art methods. [ABSTRACT FROM AUTHOR]

Since the quality of steel is of paramount importance in modern production, the defects detection of steel surface is significantly crucial. In this field, two‐stage detection algorithms have encountered issues about low detection speed, while one‐stage detection algorithms have room for improvement in detection accuracy. How to trade‐off between accuracy and speed of detection to better meet the demands of industrial production remains a challenge. To address this problem, this paper proposes a You Only Look Once version 5 (YOLOv5)‐based improved method ACD‐YOLO. ACD‐YOLO model incorporates anchors optimization, context augmentation module, and efficient convolution operators. In anchor optimization, the boundaries of anchors are optimized using an improved genetic algorithm. Moreover, to improve detection accuracy, a context augmentation module is incorporated into both the head and the backbone end of the network. Additionally, efficient convolution operators are adopted to address the increase of computation complexity caused by the context augmentation module. Experimental results show that ACD‐YOLO achieves mean average precision of 79.3%, with frames per second of 72. Compared to reference methods, ACD‐YOLO achieves the best balance between accuracy and speed of detection, and is more suitable for practice industrial production. [ABSTRACT FROM AUTHOR]

Shadow removal is a challenging task because the variety of shadows is influenced by surface texture and lighting. This paper proposes a dynamic alignment and illumination‐aware convolution (DAIC), which consists of a Feature Alignment Module (FAM) and a Dynamic Weight Module (DWM). FAM aligns the downsampled deep features with the original features and helps to extract the optimal local information to ensure that the object texture features are not corrupted. DWM generates weights according to different lighting variations for a better shadow removal result. The shadow removal approach is based on an image decomposition algorithm using a multi‐exposure image fusion model. Here, the shadow removal network and refinement network use U‐Net framework, and the transposed convolution operations are replaced with DAIC in the decoder part of U‐Net to improve the performance of the network. The experiments are conducted on two large shadow removal datasets, ISTD+ and SRD. Compared to state‐of‐the‐art methods, this model achieves optimal performance in terms of Root Mean Square Error (RMSE) for the non‐shadow region. It also achieves performance comparable to the state‐of‐the‐art method in terms of RMSE for the shadow region and structural similarity index measurement for the entire image. [ABSTRACT FROM AUTHOR]

Automatic profiled fibre recognition and analysis can accelerate quality inspection and contributes to the upgrade of the textile industry. However, these tasks often require significant manual effort to generate instance‐level annotations for fully supervised training. In this paper, the authors propose a weakly supervised pipeline for profiled fibre inspection using electron‐microscopic (EM) images with only image‐level annotations. It automatically identifies fibre instances and estimates shape factors to facilitate fibre quality inspection. As the core of the pipeline, the weakly supervised network (WesNet) is designed to localize hundreds of crowded fibre samples by raw patch generation and fibre sample sifting. Particularly, the composite similarity measurement integrates different patch‐wise similarities, enabling the network to distinguish fibre from background robustly. For quality inspection, the pipeline further analyzes the fibre instances, utilizing several efficient techniques to estimate the shape factors. Experiments on the real fibre electron‐microscopic images demonstrate the efficacy and efficiency of the pipeline. Results show that WesNet outperforms several supervised and weakly supervised methods, including two state‐of‐the‐art weakly supervised networks. [ABSTRACT FROM AUTHOR]

CABAC is the only entropy coding used in Versatile Video Coding (VVC). This is achieved through multiple estimators approach that provide more accurate predictions by considering different estimated probability results, but CABAC coding requires higher complexity and bit‐range accuracy than other approaches. Therefore, there is more potential to refine the performance from the perspective of bit allocation and architecture design. In this paper, a selection method is proposed to determine which estimator is recommended to dynamically perform the current entropy coding. Taking advantage of the double‐bit‐range architecture, the bits contained in the different estimators are also rearranged based on Most Probable Symbol (MPS$\mathit {MPS}$) determination and Least Probable Symbol (LPS$\mathit {LPS}$) considerations. Experimental reports in the work report that the coding time can be reduced using the proposed method and there is a slight gain in Peak Signal‐to‐Noise Ratio (PSNR) while saving some Rate‐distortion (RD) performance in bitrate. [ABSTRACT FROM AUTHOR]

Light undergoes attenuation due to scattering and refraction when propagating through aerosols. In foggy conditions, Aerosol particles in the troposphere exhibit high mobility introducing intricate non‐linear noise into images. Foggy image restoration represents an ill‐posed problem, where traditional physical models and image enhancement techniques often prove inadequate in delivering effective solutions. This paper introduces a novel deep sub‐pixel reconstruction algorithm for foggy image restoration, pioneering the application of sub‐pixel reconstruction modules to this domain. This model employs convolutional layers to extract low‐level features and dense‐connected layers for high‐level feature extraction. Furthermore, a specialized sub‐pixel reconstruction module tailored for the task of foggy image restoration is designed, with the purpose of reconstructing dehazed images from latent vectors. During training, a generative adversarial training framework is adopted, incorporating a purpose‐designed discriminator. Additionally, a fusion loss is implemented to facilitate model refinement. Quantitative and qualitative evaluation experiments conducted on synthetic and real‐world image datasets demonstrate the effectiveness of the proposed method in preserving finer details. The Structural Similarity Index (SSIM) is observed to improve by 2.5%, attesting to enhanced perceptual quality for grayscale foggy images. [ABSTRACT FROM AUTHOR]

When using untrusted third parties to compress and transmit images in real‐life scenarios, it is vital to encrypt them before compression. In order to better address the issues of low security in the original image and poor reconstruction quality of the encrypted image during compressed sensing, this paper proposes a 2D compressed sensing scheme for encrypted images based on complex‐valued measurement matrix (2DCS‐CVM). Firstly, the SHA‐256 algorithm generates keys for the hyperchaotic Lorenz system, and then the chaotic sequences are used to create encrypted images with increased security through subtractive diffusion and global permutation. Secondly, the complex‐valued Vandermonde measurement matrix is utilized for 2D compressed sensing on the encrypted image, and the two‐dimensional projected gradient with embedding decryption algorithm is used to generate recovered images with improved reconstruction performance. Finally, the measurement matrix's computational complexity and transmission bandwidth are reduced through structural sparsification with sparse random matrices. Simulation results demonstrate that this scheme offers an optimal balance between storage, computational complexity, hardware implementation, and reconstruction performance while providing excellent security and robustness. [ABSTRACT FROM AUTHOR]

Deep learning‐based‐image denoising methods have recently achieved excellent performance by learning non‐linear mapping in the spatial domain. However, these methods fail to address the noise without specific distribution because they only use features of the spatial domain. Meanwhile, existing methods that utilize features in the frequency domain fail to combine the detailed information of both domains properly for effective reconstruction and demonstrate poor generalization. Therefore, a novel adaptive fusion dual‐domain network (AFDN) is introduced for single image restoration. Different from deep learning‐based methods, which operate on the spatial or dual‐domains in a certain order, the proposed AFDN combine the spatial‐domain image and corresponding frequency‐domain image as the input and use the interlacing dual‐domain module with flexible adaptability to learn the relationship between spatial and frequency domains. In experimental results, the AFDN is compared with several state‐of‐the‐art restoration methods. Quantitative results showed that the AFDN achieves enhanced effects and high index values. The code of this paper will be released at https://github.com/jzw0707/AFDN [ABSTRACT FROM AUTHOR]

Morphing attacks (MAs) pose a substantial security threat to the Automatic Border Control (ABC) system. While a few morphing attack detection (MAD) methods have been proposed, the face morphing accomplice's facial restoration has not received sufficient attention. Due to the inability to foresee the morphing factor used for a particular morphed image, selecting the appropriate de‐morphing factor becomes a challenging problem in the restoration of the accomplice's facial image. If the morphing factor cannot be chosen reasonably, achieving the desired restoration effect is difficult. Therefore, this paper presents an adaptive de‐morphing factor framework (ADFF) architecture for restoring the accomplice's facial image. By exploiting the morphed images stored in the electronic passport system and the real‐time captured criminal's images, ADFF can effectively restore the accomplice's facial image. Experimental results and analysis show that ADFF can significantly reduce the security threats of MAs on ABC. [ABSTRACT FROM AUTHOR]

The accurate segmentation of breast tumours is important for the diagnosis and treatment of breast cancer. When using the classic U‐Net, Attention‐UNet, and UNet++ to segment the tumour, there are problems of oversegmentation, incorrect segmentation, and poor edge continuity. In this paper, an effective tumour segmentation method, EfficientUNet, is proposed. EfficientUNet adopts a step‐by‐step enhancement method, combining ResNet18, a channel attention mechanism and deep supervision. ResNet18, as the encoder of the whole network, solves the problem of gradient disappearanceand improves the feature extraction ability of the model. The channel attention module makes the model more accurate in tumour edge processing. The deep supervision technology accelerates the model training and provides the convergence direction for the model. In addition, it is found that when adjusting the size of the image, the method of image filling before clipping (or zooming) is more conducive to model learning than the direct interpolation method. And a comparative experiment wasperformed on dataset B. Compared to U‐Net, Attention‐UNet and UNet++, EfficientUNet has the highest performance. Finally, the ablation experiment also indicated the effectiveness of each module in EfficientUNet. [ABSTRACT FROM AUTHOR]

In this paper, a method of stroke‐based rendering is proposed for image representation and reconstruction. The proposed method involves compositing a set of ellipses that greatly vary in location, size, rotation angle, color, and opacity. This study proves that alpha compositing is differentiable if a two‐dimensional Gaussian function is used to draw a solid ellipse. The gradient method can then be employed for automatically identifying the parameters of each ellipse such that the difference between the input image and the composited image is minimized. Experimental results indicate that the proposed method can represent various types of images including portraits, landscapes, buildings, street scenes, artificial objects, medical images etc. The proposed method can particularly render the most details and significant features of an input image with fewer strokes compared to other stroke‐based rendering algorithms. This study also demonstrates that the proposed method can be applied in painting style simulation and sparse‐view computed tomography imaging. [ABSTRACT FROM AUTHOR]