Feature Preservation Research Articles

Point Cloud Denoising and Feature Preservation: An Adaptive Kernel Approach Based on Local Density and Global Statistics.

Noise removal is a critical stage in the preprocessing of point clouds, exerting a significant impact on subsequent processes such as point cloud classification, segmentation, feature extraction, and 3D reconstruction. The exploration of methods capable of adapting to and effectively handling the noise in point clouds from real-world outdoor scenes remains an open and practically significant issue. Addressing this issue, this study proposes an adaptive kernel approach based on local density and global statistics (AKA-LDGS). This method constructs the overall framework for point cloud denoising using Bayesian estimation theory. It dynamically sets the prior probabilities of real and noise points according to the spatial function relationship, which varies with the distance from the points to the center of the LiDAR. The probability density function (PDF) for real points is constructed using a multivariate Gaussian distribution, while the PDF for noise points is established using a data-driven, non-parametric adaptive kernel density estimation (KDE) approach. Experimental results demonstrate that this method can effectively remove noise from point clouds in real-world outdoor scenes while maintaining the overall structural features of the point cloud.

Variational sparse diffusion and its application in mesh processing

PurposeThe primary objective of this study is to tackle the enduring challenge of preserving feature integrity during the manipulation of geometric data in computer graphics. Our work aims to introduce and validate a variational sparse diffusion model that enhances the capability to maintain the definition of sharp features within meshes throughout complex processing tasks such as segmentation and repair.Design/methodology/approachWe developed a variational sparse diffusion model that integrates a high-order L1 regularization framework with Dirichlet boundary constraints, specifically designed to preserve edge definition. This model employs an innovative vertex updating strategy that optimizes the quality of mesh repairs. We leverage the augmented Lagrangian method to address the computational challenges inherent in this approach, enabling effective management of the trade-off between diffusion strength and feature preservation. Our methodology involves a detailed analysis of segmentation and repair processes, focusing on maintaining the acuity of features on triangulated surfaces.FindingsOur findings indicate that the proposed variational sparse diffusion model significantly outperforms traditional smooth diffusion methods in preserving sharp features during mesh processing. The model ensures the delineation of clear boundaries in mesh segmentation and achieves high-fidelity restoration of deteriorated meshes in repair tasks. The innovative vertex updating strategy within the model contributes to enhanced mesh quality post-repair. Empirical evaluations demonstrate that our approach maintains the integrity of original, sharp features more effectively, especially in complex geometries with intricate detail.Originality/valueThe originality of this research lies in the novel application of a high-order L1 regularization framework to the field of mesh processing, a method not conventionally applied in this context. The value of our work is in providing a robust solution to the problem of feature degradation during the mesh manipulation process. Our model’s unique vertex updating strategy and the use of the augmented Lagrangian method for optimization are distinctive contributions that enhance the state-of-the-art in geometry processing. The empirical success of our model in preserving features during mesh segmentation and repair presents an advancement in computer graphics, offering practical benefits to both academic research and industry applications.

Face De-Identification Using Convolutional Neural Network (CNN) Models for Visual-Copy Detection

The proliferation of media-sharing platforms has led to issues with illegally edited content and the distribution of pornography. To protect personal information, de-identification technologies are being developed to prevent facial identification. Existing de-identification methods directly alter the pixel values in the face region, leading to reduced feature representation and identification accuracy. This study aims to develop a method that minimizes the possibility of personal identification while effectively preserving important features for image- and video-copy-detection tasks, proposing a new deep-learning-based de-identification approach that surpasses traditional pixel-based alteration methods. We introduce two de-identification models using different approaches: one emphasizing the contours of the original face through feature inversion and the other generating a blurred version of the face using D2GAN (Dual Discriminator Generative Adversarial Network). Both models were evaluated on their performance in image- and video-copy-detection tasks before and after de-identification, demonstrating effective feature preservation. This research presents new possibilities for personal-information protection and digital-content security, contributing to digital-rights management and law enforcement.

Physical feature preserving and unconditionally stable SAV fully discrete finite element schemes for incompressible flows with variable density

In this paper, we construct new positive preserving, unconditionally stable and fully discrete finite element schemes for incompressible flows with variable density. The proposed schemes employ the positive function transform ρ=F(ϕ) for the density equation and scalar auxiliary variable (SAV) q=exp(−t/T) for the momentum equation in its reformulation system. The SAV schemes are unconditionally energy stable and second-order accurate in time and lead to two decoupled generalized Stokes equations for the momentum equation to be solved at each time step. Thus, it is easy to implement and extremely efficient for these schemes when combined with an adaptive time stepping method. Numerical experiments demonstrate the stability of computations and verify the second-order accuracy of the proposed schemes.

MSFANet: multi-scale fusion attention network for mangrove remote sensing lmage segmentation using pattern recognition

Mangroves are ecosystems that grow in the intertidal areas of coastal zones, playing crucial ecological roles and possessing unique economic and social values. They have garnered significant attention and research interest. Semantic segmentation of mangroves is a fundamental step for further investigations. However, mangrove remote sensing images often have large dimensions, with a substantial portion of the image containing mangrove features. Deep learning convolutional kernels may lead to inadequate receptive fields for accurate mangrove recognition. In mangrove remote sensing images, various challenges arise, including the presence of small and intricate details aside from the mangrove regions, which intensify the segmentation complexity. To address these issues, this paper primarily focuses on two key aspects: first, the exploration of methods to achieve a large receptive field, and second, the fusion of multi-scale information. To this end, we propose the Multi-Scale Fusion Attention Network (MSFANet), which incorporates a multi-scale network structure with a large receptive field for feature fusion. We emphasize preserving spatial information by integrating spatial data across different scales, employing separable convolutions to reduce computational complexity. Additionally, we introduce an Attention Fusion Module (AFM). This module helps mitigate the influence of irrelevant information and enhances segmentation quality. To retain more semantic information, this paper introduces a dual channel approach for information extraction through the deep structure of ResNet. We fuse features using the Feature Fusion Module (FFM) to combine both semantic and spatial information for the final output, further enhancing segmentation accuracy. In this study, a total of 230 images with dimensions of 768 pixels in width and height were selected for this experiment, with 184 images used for training and 46 images for validation. Experimental results demonstrate that our proposed method achieves excellent segmentation results on a small sample dataset of remote-sensing images, with significant practical value. This paper primarily focuses on three key aspects: the generation of mangrove datasets, the preprocessing of mangrove data, and the design and training of models. The primary contribution of this paper lies in the development of an effective approach for multi-scale information fusion and advanced feature preservation, providing a novel solution for mangrove remote sensing image segmentation tasks. The best Mean Intersection over Union (MIoU) achieved on the mangrove dataset is 86%, surpassing other existing models by a significant margin.

EnNuSegNet: Enhancing Weakly Supervised Nucleus Segmentation through Feature Preservation and Edge Refinement

Nucleus segmentation plays a crucial role in tissue pathology image analysis. Despite significant progress in cell nucleus image segmentation algorithms based on fully supervised learning, the large number and small size of cell nuclei pose a considerable challenge in terms of the substantial workload required for label annotation. This difficulty in acquiring datasets has become exceptionally challenging. This paper proposes a novel weakly supervised nucleus segmentation method that only requires point annotations of the nuclei. The technique is an encoder–decoder network which enhances the weakly supervised nucleus segmentation performance (EnNuSegNet). Firstly, we introduce the Feature Preservation Module (FPM) in both encoder and decoder, which preserves more low-level features from the shallow layers of the network during the early stages of training while enhancing the network’s expressive capability. Secondly, we incorporate a Scale-Aware Module (SAM) in the bottleneck part of the network to improve the model’s perception of cell nuclei at different scales. Lastly, we propose a training strategy for nucleus edge regression (NER), which guides the model to optimize the segmented edges during training, effectively compensating for the loss of nucleus edge information and achieving higher-quality nucleus segmentation. Experimental results on two publicly available datasets demonstrate that our proposed method outperforms state-of-the-art approaches, with improvements of 2.02%, 1.41%, and 1.59% in terms of F1 score, Dice coefficient, and Average Jaccard Index (AJI), respectively. This indicates the effectiveness of our method in improving segmentation performance.

Norest-Net: Normal Estimation Neural Network for 3-D Noisy Point Clouds.

The widely deployed ways to capture a set of unorganized points, e.g., merged laser scans, fusion of depth images, and structure-from- x , usually yield a 3-D noisy point cloud. Accurate normal estimation for the noisy point cloud makes a crucial contribution to the success of various applications. However, the existing normal estimation wisdoms strive to meet a conflicting goal of simultaneously performing normal filtering and preserving surface features, which inevitably leads to inaccurate estimation results. We propose a normal estimation neural network (Norest-Net), which regards normal filtering and feature preservation as two separate tasks, so that each one is specialized rather than traded off. For full noise removal, we present a normal filtering network (NF-Net) branch by learning from the noisy height map descriptor (HMD) of each point to the ground-truth (GT) point normal; for surface feature recovery, we construct a normal refinement network (NR-Net) branch by learning from the bilaterally defiltered point normal descriptor (B-DPND) to the GT point normal. Moreover, NR-Net is detachable to be incorporated into the existing normal estimation methods to boost their performances. Norest-Net shows clear improvements over the state of the arts in both feature preservation and noise robustness on synthetic and real-world captured point clouds.

Feature Preservation and Shape Cues Assist Infrared Small Target Detection

Feature Preservation and Shape Cues Assist Infrared Small Target Detection

Effect of detector placement on joint estimation in X-ray fluorescence emission tomography.

Imaging the spatial distribution of low concentrations of metal is a growing problem of interest with applications in medical and material sciences. X-ray fluorescence emission tomography (XFET) is an emerging metal mapping imaging modality with potential sensitivity improvements and practical advantages over other methods. However, XFET detector placement must first be optimized to ensure accurate metal density quantification and adequate spatial resolution. In this work, we first use singular value decomposition of the imaging model and eigendecomposition of the object-specific Fisher information matrix to study how detector arrangement affects spatial resolution and feature preservation. We then perform joint image reconstructions of a numerical gold phantom. For this phantom, we show that two parallel detectors provide metal quantification with similar accuracy to four detectors, despite the resulting anisotropic spatial resolution in the attenuation map estimate. Two orthogonal detectors provide improved spatial resolution along one axis, but underestimate the metal concentration in distant regions. Therefore, this work demonstrates the minor effect of using fewer, but strategically placed, detectors in the case where detector placement is restricted. This work is a critical investigation into the limitations and capabilities of XFET prior to its translation to preclinical and benchtop uses.

Federated Discriminative Representation Learning for Image Classification.

Acquiring big-size datasets to raise the performance of deep models has become one of the most critical problems in representation learning (RL) techniques, which is the core potential of the emerging paradigm of federated learning (FL). However, most current FL models concentrate on seeking an identical model for isolated clients and thus fail to make full use of the data specificity between clients. To enhance the classification performance of each client, this study introduces the FDRL, a federated discriminative RL model, by partitioning the data features of each client into a global subspace and a local subspace. More specifically, FDRL learns the global representation for federated communication between those isolated clients, which is to capture common features from all protected datasets via model sharing, and local representations for personalization in each client, which is to preserve specific features of clients via model differentiating. Toward this goal, FDRL in each client trains a shared submodel for federated communication and, meanwhile, a not-shared submodel for locality preservation, in which the two models partition client-feature space by maximizing their differences, followed by a linear model fed with combined features for image classification. The proposed model is implemented with neural networks and optimized in an iterative manner between the server of computing the global model and the clients of learning the local classifiers. Thanks to the powerful capability of local feature preservation, FDRL leads to more discriminative data representations than the compared FL models. Experimental results on public datasets demonstrate that our FDRL benefits from the subspace partition and achieves better performance on federated image classification than the state-of-the-art FL models.

FPCS: Feature Preserving Compensated Sampling of Streaming Time Series Data.

Data visualization aids in making data analysis more intuitive and in-depth, with widespread applications in fields such as biology, finance, and medicine. For massive and continuously growing streaming time series data, these data are typically visualized in the form of line charts, but the data transmission puts significant pressure on the network, leading to visualization lag or even failure to render completely. This paper proposes a universal sampling algorithm FPCS, which retains feature points from continuously received streaming time series data, compensates for the frequent fluctuating feature points, and aims to achieve efficient visualization. This algorithm bridges the gap in sampling for streaming time series data. The algorithm has several advantages: (1) It optimizes the sampling results by compensating for fewer feature points, retaining the visualization features of the original data very well, ensuring high-quality sampled data; (2) The execution time is the shortest compared to similar existing algorithms; (3) It has an almost negligible space overhead; (4) The data sampling process does not depend on the overall data; (5) This algorithm can be applied to infinite streaming data and finite static data.

Neural Network-Based Pipeline With High-Resolution Feature Enhancement and Low-Resolution Feature Preservation for Automated Treatment Decision of Graves’ Orbitopathy Patients

Neural Network-Based Pipeline With High-Resolution Feature Enhancement and Low-Resolution Feature Preservation for Automated Treatment Decision of Graves’ Orbitopathy Patients

A variational model to remove multiplicative noise based on SAR image feature preservation

A variational model to remove multiplicative noise based on SAR image feature preservation

영어 무성치경폐쇄음의 기식음화와 성문음화의 자질 보존성 관계 연구

This study is to argue that the choice between aspiration and glottalization of /t/ by speakers in the same phonological environment can be influenced by differences in the ranking of feature preservation constraints. Besides, I propose that if Korean learners of English tend to produce aspiration in environments where glottalization are prevalent, it results from applying the constraint favoring listener's perceptive ease being ranked higher than that favoring speaker's articulatory ease. Allophones of /t/ can vary even among native speakers due to different sociolinguistic factors. When Korean learners of English apply the constraint of [aspirated] feature preservation at a high ranking, with the intention of enhancing listener's perception, pronunciation errors may occur. Also, if they rank high the constraint for avoiding consonant cluster, which is not allowed in Korean language, they can produce awkward pronunciation. Therefore, by analyzing aspiration and glottalization of /t/ through Optimality Theory, this paper aims to demonstrate that the reason English /t/ exhibits more phonological variants and allophones is attributed to the low feature preservation it holds.

Analyzing CT Scan Images towards the Early Detection of Lung Cancer using Medical Images based Edge Feature Preserving CT Scan Medical Image Coder (EZWT - EFPIC)

With the current improvements in virtual image processing techniques have received several benefits. Today, all of the scientific techniques produce virtual scientific pictures, through healthcare specialists analyze and diagnose the abnormality. The frequent view of scientific picture processing might also additionally appear simpler; however, it entails many challenges. As the scientific pics are interconnected with human lives, the laptop-aided scientific image processing structures have to be overcautious, if we want to eliminate inaccuracy rates. The utility of medical image processing techniques for the analysis of CT scan images similar to lung cancer cells has been gaining momentum in recent years. This paper discusses the use of a Transform Edge Feature preserving CT scan Medical Image Coder (EZWT - EFPIC) using Computed Tomography (CT) images to help in the early diagnosis of lung cancer. We discuss and explore the design and significance of an EZWT-EFPIC-CT image-processed model in cancer diagnosis.

Analysing detail preserving capabilities of bilateral, laplacian and taubin mesh filtering methods

Mesh filtering of surfaces is crucial for noise reduction, feature preservation, and mesh simplification in graphics, visualization, and computer vision. In this paper, the detail preservation capacities of 3 frequently used filters, i.e., Bilateral, Laplacian, and Taubin mesh filters, in mesh filtering have been thoroughly examined by experiments conducted on 4 different test meshes. While the Bilateral filter excels in preserving sharp features due to its integration of geometric proximity with intensity similarity, the Laplacian filter prioritizes smoothness by averaging neighboring vertex positions, and the Taubin filter offers a balanced approach by merging attributes of both Laplacian and high-pass filters. The Bilateral filter's primary strength lies in its ability to maintain sharp features on a mesh, ensuring that intricate details are preserved by considering both the spatial closeness and intensity similarity of vertices. The Laplacian filter, although effective in achieving mesh smoothness, has the propensity to excessively smooth out sharp and defining features, potentially causing a loss of critical details in the processed mesh. The Taubin filter integrates the best of both worlds, ensuring smoothness without excessive mesh shrinkage; however, it might not excel in feature preservation as effectively as the Bilateral filter or smooth as uniformly as the Laplacian filter, making it a middle-ground option for certain applications. The statistical analysis of the experimental results has shown that the Taubin method is statistically a more successful mesh filtering method for the test sets used in this paper.

Classification of hyperspectral image by preprocessing method based relation network

ABSTRACTHyperspectral image classification involves assigning a land cover class to each pixel of a hyperspectral image (HSI). Recently, several classification methods based on deep learning with spatial information from HSI have been proposed. These methods typically demonstrate strong classification capabilities. However, spatial information often comprises complex data, making it challenging to extract valuable information. To address this issue, this paper proposes a novel image preprocessing method for classification based on metric learning. This method aims to reduce the complexity of spatial information to facilitate network learning. First, the method assesses the similarity between the central pixel and other pixels in each image patch using a relation network. It replaces pixels that are dissimilar to the central pixel with pixels that are similar to the central pixel to simplify the image patch. This approach reduces interference information in the image patch, allowing the network to focus more on the central pixel features that represent category information during the classification task. This greatly reduces the complexity of the required network, the difficulty of the network learning, and the number of training samples needed. Experimental results on three popular hyperspectral image datasets demonstrate that the proposed preprocessing method significantly improves classification accuracy compared to the baseline network. The study also compares our method to the state-of-the-art attention mechanism CNN, demonstrating how our method excels at hyperspectral classification in data-smoothing and feature preservation.

POINT CLOUD SIMPLIFICATION METHOD BASED ON IMPROVED FUZZY C-MEANS WITH AUTOMATIC NUMBER OF CLUSTERS

The essence of the point cloud is to express geometric information about objects by getting discrete coordinates on their surfaces. However, with a million points, this data may record redundant details in which it is needless to be kept for the model’s analysis. In addition, there is also a limitation where the processors cannot process the large-size datasets. The point cloud simplification algorithm was developed to solve the stated obstacles in data processing. Numerous algorithms have been published to produce the best methods for data reduction process. Since the simplification process might eliminate essential features of the data, this study introduces the features preservation process to keep the important points before the simplification. This study employed the Fuzzy C-Means (FCM) algorithms for the simplification stage due to their simplicity and ability to generate an accurate result. Regardless, the FCM still suffers from drawbacks, where their initial cluster centres are prone to fall into local optima. This study improved the FCM by employing the Score and Minimum Distance (SMD) to determine the number of clusters and cluster centres. The SMD is enhanced by changing the Gaussian to Cosine kernel function to increase the accuracy. This new technique is named SMD(C)-IFCM. The method was then applied to the 3D point cloud of a box, cup, and Stanford bunny. The performance of the developed method was compared with the original SMD-FCM and SMD-IFCM for the percentage of the simplified data, error evaluation, and processing time. The result and analysis showed that the developed method had the best score, which was six (6) out of nine (9) measurements, compared to the other two methods with scores of one (1) and two (2) respectively. This score suggests that the developed algorithm successfully reduced the error evaluation and the processing time to generate the output.

Triple‐loss driven generative adversarial network for pansharpening

AbstractPansharpening aims at fusing a panchromatic (PAN) image and a low‐resolution multispectral (LRMS) image into a high‐resolution multispectral (HRMS) image. In recent years, GAN‐based pansharpening methods have achieved excellent results, but they suffer from inadequate feature preservation and unstable training. To address these issues, a novel GAN‐based model named TriLossGAN is proposed. This method constructs three loss components with the help of the generator and the dual‐discriminator, which are calculated in both the original spatial domain and the transform domain to better preserve high‐frequency and low‐frequency information in the fused image. Additionally, a new training strategy is designed to stabilize the training process. In extensive experiments, the proposed method achieved satisfactory results on three datasets with QNR values of 0.9584 on GaoFen‐2, 0.9601 on QuickBird, and 0.9138 on WorldView‐3. Qualitative and quantitative comparisons demonstrate that TriLossGAN outperforms other state‐of‐the‐art methods.

Reducing the risk of hallucinations with interpretable deep learning models for low-dose CT denoising: comparative performance analysis

Objective. Reducing CT radiation dose is an often proposed measure to enhance patient safety, which, however results in increased image noise, translating into degradation of clinical image quality. Several deep learning methods have been proposed for low-dose CT (LDCT) denoising. The high risks posed by possible hallucinations in clinical images necessitate methods which aid the interpretation of deep learning networks. In this study, we aim to use qualitative reader studies and quantitative radiomics studies to assess the perceived quality, signal preservation and statistical feature preservation of LDCT volumes denoised by deep learning. We aim to compare interpretable deep learning methods with classical deep neural networks in clinical denoising performance. Approach. We conducted an image quality analysis study to assess the image quality of the denoised volumes based on four criteria to assess the perceived image quality. We subsequently conduct a lesion detection/segmentation study to assess the impact of denoising on signal detectability. Finally, a radiomic analysis study was performed to observe the quantitative and statistical similarity of the denoised images to standard dose CT (SDCT) images. Main results. The use of specific deep learning based algorithms generate denoised volumes which are qualitatively inferior to SDCT volumes(p < 0.05). Contrary to previous literature, denoising the volumes did not reduce the accuracy of the segmentation (p > 0.05). The denoised volumes, in most cases, generated radiomics features which were statistically similar to those generated from SDCT volumes (p > 0.05). Significance. Our results show that the denoised volumes have a lower perceived quality than SDCT volumes. Noise and denoising do not significantly affect detectability of the abdominal lesions. Denoised volumes also contain statistically identical features to SDCT volumes.