Search

Deep learning techniques offer improvements in computer-aided diagnosis systems. However, acquiring image domain annotations is challenging due to the knowledge and commitment required of expert pathologists. Pathologists often identify regions in whole slide images with diagnostic relevance rather than examining the entire slide, with a positive correlation between the time spent on these critical image regions and diagnostic accuracy. In this paper, a heatmap is generated to represent pathologists viewing patterns during diagnosis and used to guide a deep learning architecture during training. The proposed system outperforms traditional approaches based on color and texture image characteristics, integrating pathologists domain expertise to enhance region of interest detection without needing individual case annotations. Evaluating our best model, a U-Net model with a pre-trained ResNet-18 encoder, on a skin biopsy whole slide image dataset for melanoma diagnosis, shows its potential in detecting regions of interest, surpassing conventional methods with an increase of 20%, 11%, 22%, and 12% in precision, recall, F1-score, and Intersection over Union, respectively. In a clinical evaluation, three dermatopathologists agreed on the models effectiveness in replicating pathologists diagnostic viewing behavior and accurately identifying critical regions. Finally, our study demonstrates that incorporating heatmaps as supplementary signals can enhance the performance of computer-aided diagnosis systems. Without the availability of eye tracking data, identifying precise focus areas is challenging, but our approach shows promise in assisting pathologists in improving diagnostic accuracy and efficiency, streamlining annotation processes, and aiding the training of new pathologists.

To enhance the regional feature preserving ability and guaranteed inclusion of complete objects in the final fused images, this paper proposes a new image fusion method based on visual saliency objects and fuzzy region attributes. In this approach, the complementary feature encoders are used to extract the features of infrared and visible images, respectively. In the feature fusion layer, to avoid the loss of target features caused by pixel-level fusion and region rigid division, this paper introduces the fuzzy set theory and salient target detection into the design of the fusion strategy. Following the k-means clustering and the energy scale of the infrared scene, the essential attributes of the areas are established. Meanwhile, a salient object segmentation algorithm based on a convolution neural network is adopted to extract the salient region of the source image. Then, feature maps are fused using fuzzy region memberships and salient feature mapping. Finally, the fused image is reconstructed by a redundant feature decoder. The proposed method focuses on the fusion of color scene images and multi-wavelength infrared (thermal and near-infrared) images. Experiments are performed on public RGB-T and RGB-NIR datasets. Compared with several start-of-the-art methods, the evaluation of visual and objective metrics demonstrates that the proposed fusion method has superior performance and is more in line with human sensory characteristics. [ABSTRACT FROM AUTHOR]

To effectively preserve thermal targets in infrared (IR) images and texture information in visible (VIS) images, this paper proposes a saliency detection-based and multi-scale oriented features image fusion method. Firstly, an edge guidance network is used to extract salient targets from IR images in order to generate salient target masks, which provide direction for fusing various types of information and increase the network’s generalization capability. Secondly, a particular loss function in each region is developed in conjunction with the salient target masks to direct the network to perform feature extraction. It can extract salient target characteristics and background texture information selectively while maintaining the integrity of the salient target and background regions efficiently. Finally, an oriented fusion approach based on feature hierarchy is proposed, which minimizes information dropout by integrating deep features on distinct scales with direction. Extensive experiments are carried out in this research using available datasets such as TNO and RoadScene. The method outperforms 10 state-of-the-art methods in the most of the evaluation metrics. This demonstrates that proposed fusion method not only preserves the salient target information of IR images, but also retains more VIS image features. [ABSTRACT FROM AUTHOR]

Gestalt theory laid the foundation for modern cognitive learning theory and emphasizes that the whole is greater than the sum of its parts, where similarity and proximity are two important principles. However, exploiting Gestalt theory to detect multiple salient objects remains challenging. In this paper, we propose a very simple yet efficient saliency model based on Gestalt theory, namely, the color similarity and spatial proximity (CSSP) model. It utilizes content-based image retrieval (CBIR) techniques to detect salient objects. The methodology has three important highlights: (1) a novel weighted distance is proposed to calculate spatial proximity. It can control spatial proximity within a certain range and detect salient objects robustly. (2) Two novel and efficient saliency scoring calculation methods are proposed under the framework of CBIR techniques, where color similarity and spatial proximity are used for image matching and the ordering of retrieved images. This enables the robust identification of multiple salient objects. (3) A very simple yet efficient integration method is proposed to combine saliency maps. Using this integration method, impurities around salient objects are greatly reduced, and their interiors are highlighted robustly. Experiments with several well-known benchmark datasets validate the performance of the CSSP model. The CSSP method resulted in fewer grey patches inside salient objects, and it is superior to many existing state-of-the-art methods. The detected salient regions were brighter, improving the effectiveness of multiple salient objects detection. In addition, the CSSP method can detect salient objects robustly even when they touch the image boundaries. It has demonstrated that modeling visual attention based on Gestalt theory is a novel, viable approach. [ABSTRACT FROM AUTHOR]

Abstract Light field imaging contains abundant scene structure information, which can improve the accuracy of salient object detection in challenging tasks and has received widespread attention. However, how to apply the abundant information of light field imaging to salient object detection still faces enormous challenges. In this paper, the lightweight attention and residual convLSTM network is proposed to address this issue, which is mainly composed of the lightweight attention-based feature enhancement module (LFM) and residual convLSTM-based feature integration module (RFM). The LFM can provide an attention map for each focal slice through the attention mechanism to focus on the features related to the object, thereby enhancing saliency features. The RFM leverages the residual mechanism and convLSTM to fully utilize the spatial structural information of focal slices, thereby achieving high-precision feature fusion. Experimental results on three publicly available light field datasets show that the proposed method surpasses the existing 17 state-of-the-art methods and achieves the highest score among five quantitative indicators.

Background: Co-salient object detection (Co-SOD) aims to identify and segment commonly salient objects in a set of related images. However, most current Co-SOD methods encounter issues with the inclusion of irrelevant information in the co-representation. These issues hamper their ability to locate co-salient objects and significantly restrict the accuracy of detection. Methods: To address this issue, this study introduces a novel Co-SOD method with iterative purification and predictive optimization (IPPO) comprising a common salient purification module (CSPM), predictive optimizing module (POM), and diminishing mixed enhancement block (DMEB). Results: These components are designed to explore noise-free joint representations, assist the model in enhancing the quality of the final prediction results, and significantly improve the performance of the Co-SOD algorithm. Furthermore, through a comprehensive evaluation of IPPO and state-of-the-art algorithms focusing on the roles of CSPM, POM, and DMEB, our experiments confirmed that these components are pivotal in enhancing the performance of the model, substantiating the significant advancements of our method over existing benchmarks. Experiments on several challenging benchmark co-saliency datasets demonstrate that the proposed IPPO achieves state-of-the-art performance.

Introduction: RGB-T Salient Object Detection (SOD) aims to accurately segment salient regions in both visible light and thermal infrared images. However, many existing methods overlook the critical complementarity between these modalities, which can enhance detection accuracy. Methods: We propose the Edge-Guided Feature Fusion Network (EGFF-Net), which consists of cross-modal feature extraction, edge-guided feature fusion, and salience map prediction. Firstly, the cross-modal feature extraction module captures and aggregates united and intersecting information in each local region of RGB and thermal images. Then, the edge-guided feature fusion module enhances the edge features of salient regions, considering that edge information is very helpful in refining significant area details. Moreover, a layer-by-layer decoding structure integrates multi-level features and generates the prediction of salience maps. Results: We conduct extensive experiments on three benchmark datasets and compare EGFF-Net with state-of-the-art methods. Our approach achieves superior performance, demonstrating the effectiveness of the proposed modules in improving both detection accuracy and boundary refinement. Discussion: The results highlight the importance of integrating cross-modal information and edge-guided fusion in RGB-T SOD. Our method outperforms existing techniques and provides a robust framework for future developments in multi-modal saliency detection. [ABSTRACT FROM AUTHOR]

Object detection algorithms require a large amount of annotated data for training and optimization, which can be time-consuming, expensive, and limit model robustness and generalization. The natural world has a diverse range of target categories, and privacy and security concerns make obtaining accurate data a challenge. Researchers are developing new methods and technologies to achieve fast and accurate object detection with minimal annotations. Recent few-shot object detection methods have employed semi-supervised learning, domain adaptation, and meta-learning techniques to enable efficient knowledge transfer from base to new categories. However, these methods have not directly addressed the primary challenge of few-shot object detection, which is the lack of sufficient labeled data for new categories. This study introduces SAPT, a few-shot object detection method based on saliency data augmentation and unsupervised pre-training model fusion. In the data preprocessing stage, SAPT selects two images for detection and crops an object of the same category from each based on ground truth labels. The maximum and minimum salient circular regions of the two cropped objects are detected and blended to generate an augmented image that expands the training dataset. In the testing stage, the output of the supervised few-shot object detection and unsupervised pre-training models are integrated and mined to generate dynamic positive and negative support images, improving the detector's accuracy. SAPT optimizes the detection process by addressing issues of insufficient data for unknown class labels and low detection accuracy in open-world scenarios. Extensive experiments on benchmark datasets MS COCO and PASCAL VOC demonstrate SAPT's superior performance compared to existing few-shot object detection methods. [ABSTRACT FROM AUTHOR]

In recent decades, the advancement of deep learning algorithms and their effectiveness in saliency detection has garnered significant attention in research. Among these methods, U Network ( U-Net ) is widely used in computer vision and image processing. However, most previous deep learning-based saliency detection methods have focused on the accuracy of salient regions, often overlooking the quality of boundaries, especially fine boundaries. To address this gap, we developed a method to detect boundaries effectively. This method comprises two modules: prediction and residual refinement, based on U-Net structure. The refinement module improves the mask predicted by the prediction module. Additionally, to boost the refinement of the saliency map, a channel attention module is integrated. This module has a significant impact on our proposed method. The channel attention module is implemented in the refinement module, aiding our network in obtaining a more accurate estimation by focusing on the crucial and informative regions of the image. To evaluate the developed method, five well-known saliency detection datasets are employed. The proposed method consistently outperforms the baseline method across all five datasets, demonstrating improved performance. [ABSTRACT FROM AUTHOR]

Learning novel object concepts from limited samples remains a considerable challenge in deep learning. The main directions for improving the few-shot learning models include (i) designing a stronger backbone, (ii) designing a powerful (dynamic) meta-classifier, and (iii) using a larger pre-training set obtained by generating or hallucinating additional samples from the small scale dataset. In this paper, we focus on item (iii) and present a novel meta-hallucination strategy. Presently, most image generators are based on a generative network (i.e., GAN) that generates new samples from the captured distribution of images. However, such networks require numerous annotated samples for training. In contrast, we propose a novel saliency-based end-to-end meta-hallucinator, where a saliency detector produces foregrounds and backgrounds of support images. Such images are fed into a two-stream network to hallucinate feature samples directly in the feature space by mixing foreground and background feature samples. Then, we propose several novel mixing strategies that improve the quality and diversity of hallucinated feature samples. Moreover, as not all saliency maps are meaningful or high quality, we further introduce a meta-hallucination controller that decides which foreground feature samples should participate in mixing with backgrounds. To our knowledge, we are the first to leverage saliency detection for few-shot learning. Our proposed network achieves state-of-the-art results on publicly available few-shot image classification and anomaly detection benchmarks, and outperforms competing sample mixing strategies such as the so-called Manifold Mixup. [ABSTRACT FROM AUTHOR]

Rich heterogeneous data provided by social networks can be very big, which imposes considerable challenges for object extraction and image classification. Therefore, the objective of this work is to propose an unsupervised object cosegmentation method that could be notably efficient to improve image classification performance. The main goal of cosegmentation is to extract the salient common objects within each image. To this end, we propose to minimize an energy function based on the Markov Random Field using the saliency detection, while considering linear dependence of generated foreground histograms of the input image collection. In fact, the saliency detection is processed in two steps. In each image, we detect salient objects, by considering appearance similarity and spatial distributions of image pixels. Then, fuzzy quantification is used to correct the belonging of pixels to the foreground objects. Finally, an iterative optimization permits to enhance the final segmentation results. The proposed method has been validated as a preprocessing step for image classification. Indeed, to enhance cosegmentation-based classification performance, we have applied a semi-supervised object classification based on ensemble projection. Qualitative and quantitative evaluations of the proposed cosegmentation and classification techniques on the iCoseg, CDS and Oxford Flowers 17 datasets demonstrate the effectiveness of the proposed framework. [ABSTRACT FROM AUTHOR]

The field of computer vision has witnessed significant interest in the area of salient object recognition. The utilization of this technology becomes advantageous in various applications, such as image segmentation, image attention retargeting, image cropping, and image understanding. The primary challenges encountered by underwater images pertain to diminished contrast, distorted colors, and an overall suboptimal visual appearance. The present study introduces an innovative and resilient method for detecting saliency in underwater photos. The RGB picture input undergoes enhancement and is subsequently subjected to gradient domain filtration. The filtered image is subjected to a pyramid decomposition. The computation of saliency is performed on three distinct scales using a process consisting of two parallel phases. In the first stage, saliency is calculated by employing cellular automata at various scales. Initially, the filtered image is subjected to a decomposition process into superpixels. Subsequently, the k-means clustering technique is employed to calculate matrices that represent color dissimilarity and geodesic distance. The application of cellular automata follows the fusion of the matrices. The map undergoes additional optimization and filtering processes in order to achieve saliency at a specific scale. The computation of a main saliency map involves the process of scale integration. In the second stage, the computation of saliency is achieved by applying a homomorphic filter to each scale. The natural logarithm of the filtered image is computed. The spatial domain of the image undergoes a transformation to the frequency domain. The Butterworth high-pass filter is utilized. The frequency domain of the image is subsequently converted back to its spatial domain. The computation of the inverse logarithm of the image is performed. The guided filter is employed to acquire saliency on a specific scale. The computation of the secondary saliency map involves the process of scale integration. The final output is obtained by applying multiplicative fusion to both primary and secondary saliency maps. Utilizing cutting-edge methodologies, a thorough evaluation that includes both qualitative and quantitative analyses evaluates the effectiveness of the suggested strategy. In comparison to alternative state-of-the-art methodologies, the findings indicate that the suggested methodology exhibits high levels of precision and dependability. [ABSTRACT FROM AUTHOR]

Due to their great balance between excellent performance and high efficiency, discriminative correlation filter (DCF) tracking methods for unmanned aerial vehicles (UAVs) have gained much attention. Due to these correlations being capable of being efficiently computed in a Fourier domain by discrete Fourier transform (DFT), the DFT of an image has symmetry in the Fourier domain. However, DCF tracking methods easily generate unwanted boundary effects where the tracking object suffers from challenging situations, such as deformation, fast motion and occlusion. To tackle the above issue, this work proposes a novel saliency-aware and spatial–temporal regularized correlation filter (SSTCF) model for visual object tracking. First, the introduced spatial–temporal regularization helps build a more robust correlation filter (CF) and improve the temporal continuity and consistency of the model to effectively lower boundary effects and enhance tracking performance. In addition, the relevant objective function can be optimized into three closed-form subproblems which can be addressed by using the alternating direction method of multipliers (ADMM) competently. Furthermore, utilizing a saliency detection method to acquire a saliency-aware weight enables the tracker to adjust to variations in appearance and mitigate disturbances from the surroundings environment. Finally, we conducted numerous experiments based on three different benchmarks, and the results showed that our proposed model had better performance and higher efficiency compared to the most advanced trackers. For example, the distance precision (DP) score was 0.883, and the area under the curve (AUC) score was 0.676 on the OTB2015 dataset. [ABSTRACT FROM AUTHOR]

LiDAR odometry estimation and 3D semantic segmentation are crucial for autonomous driving, which has achieved remarkable advances recently. However, these tasks are challenging due to the imbalance of points in different semantic categories for 3D semantic segmentation and the influence of dynamic objects for LiDAR odometry estimation, which increases the importance of using representative/salient landmarks as reference points for robust feature learning. To address these challenges, we propose a saliency-guided approach that leverages attention information to improve the performance of LiDAR odometry estimation and semantic segmentation models. Unlike in the image domain, only a few studies have addressed point cloud saliency information due to the lack of annotated training data. To alleviate this, we first present a universal framework to transfer saliency distribution knowledge from color images to point clouds, and use this to construct a pseudo-saliency dataset (i.e. FordSaliency) for point clouds. Then, we adopt point cloud based backbones to learn saliency distribution from pseudo-saliency labels, which is followed by our proposed SalLiDAR module. SalLiDAR is a saliency-guided 3D semantic segmentation model that integrates saliency information to improve segmentation performance. Finally, we introduce SalLONet, a self-supervised saliency-guided LiDAR odometry network that uses the semantic and saliency predictions of SalLiDAR to achieve better odometry estimation. Our extensive experiments on benchmark datasets demonstrate that the proposed SalLiDAR and SalLONet models achieve state-of-the-art performance against existing methods, highlighting the effectiveness of image-to-LiDAR saliency knowledge transfer. Source code will be available at [ABSTRACT FROM AUTHOR]

IntroductionRGB-T Salient Object Detection (SOD) aims to accurately segment salient regions in both visible light and thermal infrared images. However, many existing methods overlook the critical complementarity between these modalities, which can enhance detection accuracy.MethodsWe propose the Edge-Guided Feature Fusion Network (EGFF-Net), which consists of cross-modal feature extraction, edge-guided feature fusion, and salience map prediction. Firstly, the cross-modal feature extraction module captures and aggregates united and intersecting information in each local region of RGB and thermal images. Then, the edge-guided feature fusion module enhances the edge features of salient regions, considering that edge information is very helpful in refining significant area details. Moreover, a layer-by-layer decoding structure integrates multi-level features and generates the prediction of salience maps.ResultsWe conduct extensive experiments on three benchmark datasets and compare EGFF-Net with state-of-the-art methods. Our approach achieves superior performance, demonstrating the effectiveness of the proposed modules in improving both detection accuracy and boundary refinement.DiscussionThe results highlight the importance of integrating cross-modal information and edge-guided fusion in RGB-T SOD. Our method outperforms existing techniques and provides a robust framework for future developments in multi-modal saliency detection.

Wireless image sensor networks (WISNs) are widely applied in wildlife protection as they present a better performance in remote, real-time monitoring. However, traditional WISNs suffer from the limitations of low processing capability, power consumption restrictions, and narrow transmission bandwidth, which leads to a shorter working lifetime of the monitoring system when transmitting the wildlife monitoring image with high resolution. We propose a high-efficiency distributed image compression coding method based on soft threshold iteration and quantitative perception for wildlife monitoring images to rationally assign the electricity resource. Specifically, we first utilize the histogram contrast algorithm to detect the saliency object region from the original samples and use it to generate the mask image of the wildlife region. After the mask image is obtained, the distributed image compression coding method is utilized to transmit the wildlife image, in which the saliency image region is directly transmitted as a cluster head to ensure the transmission efficiency of the wildlife region. Then the background region is assigned to the other four monitoring nodes at the same level for processing and transmission, extending the lifetime of the network. Furthermore, the soft threshold iteration algorithm is utilized to encode the image data; this is suitable for WISNs. The experimental results on our own wildlife dataset show improvements of 7.47 and 9.06% for the peak signal-tonoise ratio and 16.98 and 19.50% for the structural similarity index on the reconstructed image compared with those of the discrete cosine transform and embedded zerotree wavelets algorithms, respectively. Compared with the multihop and single-hop transmission methods, the power consumption is reduced by 29.96 and 40.84%, respectively. The results of this study indicate that the WISN technique can provide feasible solutions for the intelligent monitoring of forest biological resources. [ABSTRACT FROM AUTHOR]

Recently, RGB–T saliency detection becomes gradually a hot topic due to the fact that RGB–T multi-modal data could overcome the limitation of conventional RGB data in some cases. However, existing RGB–T saliency detection methods usually fail to take both advantages of two modalities and cannot boost performance effectively. Therefore, we achieve RGB–T saliency detection via a novel method, namely cross-modal collaborative propagation (CMCP), which contains a novel saliency propagation mechanism and a novel cross-modal collaborative learning framework relied on the proposed propagation mechanism. More specifically, we firstly propose a novel saliency propagation method and then, respectively, regard two modalities as inputs to generate RGB-induced and thermal-induced propagation mechanisms. To bridge RGB–T modalities, a novel cross-modal collaborative learning framework between RGB-induced and thermal-induced propagation mechanisms is devised to optimize, respectively, two propagation results. In other words, two modalities constantly extract supervision information to help the opposite side to refine propagation result until attaining a stable state. Finally, we integrate two modalities-induced propagation results into a refined saliency map. We compare our model with the state-of-the-art RGB–T and RGB saliency detection algorithms on three benchmark datasets, and experimental results show that the proposed CMCP achieves the significant improvement. [ABSTRACT FROM AUTHOR]

Finding the most interesting areas of an image is the aim of saliency detection. Conventional methods based on low-level features rely on biological cues like texture and color. These methods, however, have trouble with processing complicated or low-contrast images. In this paper, we introduce a deep neural network-based saliency detection method. First, using semantic segmentation, we construct a pixel-level model that gives each pixel a saliency value depending on its semantic category. Next, we create a region feature model by combining both hand-crafted and deep features, which extracts and fuses the local and global information of each superpixel region. Third, we combine the results from the previous two steps, along with the over-segmented superpixel images and the original images, to construct a multi-level feature model. We feed the model into a deep convolutional network, which generates the final saliency map by learning to integrate the macro and micro information based on the pixels and superpixels. We assess our method on five benchmark datasets and contrast it against 14 state-of-the-art saliency detection algorithms. According to the experimental results, our method performs better than the other methods in terms of F-measure, precision, recall, and runtime. Additionally, we analyze the limitations of our method and propose potential future developments. [ABSTRACT FROM AUTHOR]

Video anomaly detection and localization is the process of spatiotemporally localizing the anomalous video segment corresponding to the abnormal event or activities. It is challenging due to the inherent ambiguity of anomalies, diverse environmental factors, the intricate nature of human activities, and the absence of adequate datasets. Further, the spatial localization of the video anomalies (video anomaly localization) after the temporal localization of the video anomalies (video anomaly detection) is also a complex task. Video anomaly localization is essential for pinpointing the anomalous event or object in the spatial domain. Hence, the intelligent video surveillance system must have video anomaly detection and localization as key functionalities. However, the state-of-the-art lacks a dedicated survey of video anomaly localization. Hence, this article comprehensively surveys the cutting-edge approaches for video anomaly localization, associated threshold selection strategies, publicly available datasets, performance evaluation criteria, and open trending research challenges with potential solution strategies.

Saliency detection is increasingly a crucial task in the computer vision area. In previous graph-based saliency detection, superpixels are usually regarded as the primary processing units to enhance computational efficiency. Nevertheless, most methods do not take into account the potential impact of errors in superpixel segmentation, which may result in incorrect saliency values. To address this issue, we propose a novel approach that leverages the diversity of superpixel algorithms and constructs a multi-layer graph. Specifically, we segment the input image into multiple sets by different superpixel algorithms. Through connections within and connections between these superpixel sets, we can mitigate the errors caused by individual algorithms through collaborative solutions. In addition to spatial proximity, we also consider feature similarity in the process of graph construction. Connecting superpixels that are similar in feature space can force them to obtain consistent saliency values, thus addressing challenges brought by the scattered spatial distribution and the uneven internal appearance of salient objects. Additionally, we use the two-stage manifold ranking to compute the saliency value of each superpixel, which includes a background-based ranking and a foreground-based ranking. Finally, we employ a mean-field-based propagation method to refine the saliency map iteratively and achieve smoother results. To evaluate the performance of our approach, we compare our work with multiple advanced methods in four datasets quantitatively and qualitatively.

There is a growing demand for tea harvesting robots due to the harsh harvesting environment and rising in labor costs. Despite considerable efforts by the research community, the robustness to hilly field environment and the harvesting efficiency of the device in mechanical harvesting of bulk tea remain unimproved. To lay the foundation for automated tea harvesting, this paper proposes an integrated system of bulk tea harvesting robots with autonomous profiling logic based on computer vision. A saliency detection algorithm is applied to detect tender leaves, and a target localization system is designed with it. Moreover, a depth cue-based profiling logic and the corresponding pose adjustment strategy for cutting tool are detailed. And an actuator driven by a combination of motors is developed to provide a precise and flexible motion control to the cutting tool. Additionally, the standard position of the RGB-D camera, accuracy and timeliness of the profiling operation are confirmed by the field experiment. The results of field experiment show that the average harvesting accuracy is 87.7%, and failure rate is controlled within 15%. An analysis of failure causes reveals that damage and cutting failure are the primary reasons for the unsuccessful harvesting.

The rapid development of deep learning has made a great progress in salient object detection task. Fully supervised methods need a large number of pixel-level annotations. To avoid laborious and consuming annotation, weakly supervised methods consider low-cost annotations such as category, bounding-box, scribble, etc. Due to simple annotation and existing large-scale classification datasets, the category annotation based methods have received more attention while still suffering from inaccurate detection. In this work, we proposed one weakly supervised method with category annotation. First, we proposed one coarse object location network (COLN) to roughly locate the object of an image with category annotation. Second, we refined the coarse object location to generate pixel-level pseudo-labels and proposed one quality check strategy to select high quality pseudo labels. To this end, we studied COLN twice followed by refinement to obtain a pseudo-labels pair and calculated the consistency of pseudo-label pairs to select high quality labels. Third, we proposed one multi-decoder neural network (MDN) for saliency detection supervised by pseudo-label pairs. The loss of each decoder and between decoders are both considered. Last but not least, we proposed one pseudo-labels update strategy to iteratively optimize pseudo-labels and saliency detection models. Performance evaluation on four public datasets shows that our method outperforms other image category annotation based work.

Polarimetric SAR ship detection is an important application of the polarimetric SAR system. Existing polarimetric SAR ship detection methods are plagued by erroneous detection of strong clutter and missed detection of small targets in multiscale situations. Particularly, the existing methods easily detect strong clutter as the target under strong background clutter, resulting in false alarms; in the case of multiscale ship detection, small ships are easily submerged in background clutter, resulting in missed detection of small targets. To solve these problems, this paper proposes a polarimetric SAR ship detection method based on superpixels and sparse reconstruction saliency. This method has two stages. In the first stage, the large polarimetric SAR ship detection scene image is segmented using the superpixel segmentation method to obtain a superpixel image. With the superpixel as the basic unit, a saliency detection method based on sparse reconstruction is used to obtain the saliency value of each superpixel in the image. Then, the superpixels that may contain ship targets are retained using the sea surface ship density defined in this paper. Accordingly, in the first stage, the superpixel regions that may contain ship targets are obtained through superpixel segmentation and sparse reconstruction saliency detection. Next, in the second stage, a saliency detection method based on sparse reconstruction is used to obtain the saliency value of each pixel in these reserved superpixel regions. Finally, the global threshold segmentation method is used for the pixels in these regions to obtain the final detection results of ship targets. In this paper, two polarimetric SAR images of the ALOS-2 satellite with different scenes were selected for an experiment. One image contains strong clutter on the sea surface; the other contains ships of different sizes and many small ships. The experimental results show that the proposed method can well determine the superpixel regions that may contain ship targets in the first stage and successfully obtain the ship detection results in the second stage. In addition, in both scenarios, the classic constant false alarm rate (CFAR) methods and a saliency detection method are used for comparison with the proposed method. The experimental results show that the proposed method produces almost no false alarms because it is insensitive to strong clutter; moreover, this method rarely misses small ship targets in the multiscale ship detection scene. The figure of merit of the proposed method reaches 94.87% in the strong clutter scene and 94.05% in the multiscale ship detection scene.

• A novel architecture is proposed for frescoes inpainting. The computing-sensing fusion is adopted to realize the fusion association between the user-driven virtual restored frescoes and the real space environment. • A spatial hierarchical and consistent visual salient region detection method is proposed to determine the virtual-real fusion region. • The comparison of user study and quality analysis on object fusion task proves the superiority of proposed method have subjective interaction and excellent visual perception. In the era of artificial intelligence, image-based virtual restoration of cultural relics is one of the methods used in the restoration of cultural relics. As the most informative representative of cultural heritage in the study of historical materials, murals occupy a significant position in archaeology and ancient culture. Currently, most of the existing virtual restoration of murals is limited to the restoration of image information for local damage. The scenes of murals have large spatial scales and complex semantic contents. In order to enhance the semantic relevance of the virtual restoration of murals and the immersion of a visual perception, this paper proposes a kind of mural virtual-real fusion restoration display method based on visual attention mechanism. Based on the case study on the immersive fusion restoration of Dunhuang grotto murals, the feasibility of regional mural image restoration and real space scene fusion restoration is verified. A new paradigm of mural heritage protection in the context of virtual-real fusion is realized. [ABSTRACT FROM AUTHOR]

Road drivable area detection is an important task in computer vision with applications in self-driving cars. Accurately detecting and mapping drivable areas in a scene allow vehicles and robots to plan safe trajectories. In this paper, a novel cognitively inspired approach is proposed that considers both the salient areas in a driving scene and the driver's attention mechanism. Specifically, the attention point is computed by combining salient areas and attention regions. Furthermore, we use the attention point and two boundary nodes on the road edge to form a triangular road surface area. Finally, we segment this area and remove the salient region within this area to obtain the drivable road area. Experimental results show that our proposed method can address the shortcomings of traditional vanishing point detection algorithms and enhance drivable area perception when combined with 4 different backbones on the DeepLabV3+ model. In particular, we demonstrate the effectiveness of merging salient area and attention area algorithms and explore the joint understanding of these complementary visual cues. [ABSTRACT FROM AUTHOR]

The geometric distortion in panoramic images significantly mediates the performance of saliency detection method based on traditional CNN. The strategy of dynamically expanding convolution kernel can achieve good results, but it also produces a lot of computational overhead in the process of reading the adjacency list, which decreases the computational efficiency. The appearance of graph convolution provides a new way to solve such problems. Although using graph convolution can effectively extract the structural features of the graph, it reduces the accuracy of the model resulting from ignoring the spatial features of the image signal. To this end, this paper proposes a construction method of the multi-scale graph structure of the panoramic image and a panoramic image saliency detection model composed of an image saliency feature extraction network and multi-scale saliency feature fusion network combining the image structure information and spatial information in the panoramic image. First, we establish a graph structure consisting of root and leaf nodes obtained by super-pixel segmentation at different scales and spherical Fibonacci sampling, respectively. Then, a feature extraction network composed of two graph convolution layers and two one-dimensional auto-encoders with the same parameterization is used to extract the salient features of the multi-scale graph structure. Finally, the U-Net network fuses the multi-scale saliency features to get the final saliency map. The results show that the proposed model performs better than the state-of-the-art models in terms of calculation speed and accuracy. [ABSTRACT FROM AUTHOR]

In recent years, there has been increased interest in multispectral pedestrian detection using visible and infrared image pairs. This is due to the complementary visual information provided by these modalities, which enhances the robustness and reliability of pedestrian detection systems. However, current research in multispectral pedestrian detection faces the challenge of effectively integrating different modalities to reduce miss rates in the system. This article presents an improved method for multispectral pedestrian detection. The method utilises a saliency detection technique to modify the infrared image and obtain an infrared- enhanced map with clear pedestrian features. Subsequently, a multiscale image features fusion network is designed to efficiently fuse visible and IR-enhanced maps. Finally, the fusion network is supervised by three loss functions for illumination perception, light intensity, and texture information in conjunction with the light perception sub-network. The experimental results demonstrate that the proposed method improves the logarithmic mean miss rate for the three main subgroups (all day, day and night) to 3.12%, 3.06%, and 4.13% respectively, at "reasonable" settings. This is an improvement over the traditional method, which achieved rates of 3.11%, 2.77%, and 2.56% respectively, thus demonstrating the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In order to solve the local overexposure caused by uneven surface reflectance, this paper proposes a fast-adaptive illumination control method with a camera-projector system. At first, an image is captured by the camera and the local overexposed area is segmented using saliency detection. Then the calculated image is projected onto the object by the projector as corrective illumination. The calculation process includes the inversion of the gray value in the overexposed area and the adjustment based on the position and depth information of the object. The high-exposure saturated regional which affects the target recognition is thus reduced, and the original illumination intensity is reserved for the other regions. This process is iterated until the optimal illumination is achieved. The resulting image for each iteration is evaluated using Blind/no Reference Image Space Quality Estimator (BRISQUE). When BRISQUE value reaches the minimum, a high-quality image is achieved. The experiments show that the proposed approach can significantly improve the speed of obtaining normally exposed images, and this system provides new ideas for industry image acquisition. [ABSTRACT FROM AUTHOR]