This paper presents a corner-based image alignment algorithm based on the procedures of corner-based template matching and geometric parameter estimation. This algorithm consists of two stages: 1) training phase, and ...This paper presents a corner-based image alignment algorithm based on the procedures of corner-based template matching and geometric parameter estimation. This algorithm consists of two stages: 1) training phase, and 2) matching phase. In the training phase, a corner detection algorithm is used to extract the corners. These corners are then used to build the pyramid images. In the matching phase, the corners are obtained using the same corner detection algorithm. The similarity measure is then determined by the differences of gradient vector between the corners obtained in the template image and the inspection image, respectively. A parabolic function is further applied to evaluate the geometric relationship between the template and the inspection images. Results show that the corner-based template matching outperforms the original edge-based template matching in efficiency, and both of them are robust against non-liner light changes. The accuracy and precision of the corner-based image alignment are competitive to that of edge-based image alignment under the same environment. In practice, the proposed algorithm demonstrates its precision, efficiency and robustness in image alignment for real world applications.展开更多
Deep learning networks are increasingly exploited in the field of neuronal soma segmentation.However,annotating dataset is also an expensive and time-consuming task.Unsupervised domain adaptation is an effective metho...Deep learning networks are increasingly exploited in the field of neuronal soma segmentation.However,annotating dataset is also an expensive and time-consuming task.Unsupervised domain adaptation is an effective method to mitigate the problem,which is able to learn an adaptive segmentation model by transferring knowledge from a rich-labeled source domain.In this paper,we propose a multi-level distribution alignment-based unsupervised domain adaptation network(MDA-Net)for segmentation of 3D neuronal soma images.Distribution alignment is performed in both feature space and output space.In the feature space,features from different scales are adaptively fused to enhance the feature extraction capability for small target somata and con-strained to be domain invariant by adversarial adaptation strategy.In the output space,local discrepancy maps that can reveal the spatial structures of somata are constructed on the predicted segmentation results.Then thedistribution alignment is performed on the local discrepancies maps across domains to obtain a superior discrepancy map in the target domain,achieving refined segmentation performance of neuronal somata.Additionally,after a period of distribution align-ment procedure,a portion of target samples with high confident pseudo-labels are selected as training data,which assist in learning a more adaptive segmentation network.We verified the superiority of the proposed algorithm by comparing several domain adaptation networks on two 3D mouse brain neuronal somata datasets and one macaque brain neuronal soma dataset.展开更多
Recently, the generative adversarial network(GAN) has been extensively applied to the cross-modality conversion of medical images and has shown outstanding performance than other image conversion algorithms. Hence, we...Recently, the generative adversarial network(GAN) has been extensively applied to the cross-modality conversion of medical images and has shown outstanding performance than other image conversion algorithms. Hence, we propose a novel GAN-based multi-domain registration method named multiscale diffeomorphic jointed network of registration and synthesis(MDJRS-Net). The deviation of the generator of the GAN-based approach affects the alignment phase, so a joint training strategy is introduced to improve the performance of the generator, which feedbacks the structural loss contained in the deformation field. Meanwhile, the nature of diffeomorphism can enable the network to generate deformation fields with more anatomical properties. The average dice score(Dice) is improved by 1.95% for the computer tomography venous(CTV) to magnetic resonance imaging(MRI) registration task and by 1.92% for the CTV to computer tomography plain(CTP) task compared with the other methods.展开更多
Global motion estimation (GME) algorithms are widely applied to computer vision and video processing. In the previous works, the image resolutions are usually low for the real-time requirement (e.g. video stabilizatio...Global motion estimation (GME) algorithms are widely applied to computer vision and video processing. In the previous works, the image resolutions are usually low for the real-time requirement (e.g. video stabilization). However, in some mobile devices applications (e.g. image sequence panoramic stitching), the high resolution is necessary to obtain satisfactory quality of panoramic image. However, the computational cost will become too expensive to be suitable for the low power consumption requirement of mobile device. The full search algorithm can obtain the global minimum with extremely computational cost, while the typical fast algorithms may suffer from the local minimum problem. This paper proposed a fast algorithm to deal with 2560 × 1920 high-resolution (HR) image sequences. The proposed method estimates the motion vector by a two-level coarse-to-fine scheme which only exploits sparse reference blocks (25 blocks in this paper) in each level to determine the global motion vector, thus the computational costs are significantly decreased. In order to increase the effective search range and robustness, the predictive motion vector (PMV) technique is used in this work. By the comparisons of computational complexity, the proposed algorithm costs less addition operations than the typical Three-Step Search algorithm (TSS) for estimating the global motion of the HR images without the local minimum problem. The quantitative evaluations show that our method is comparable to the full search algorithm (FSA) which is considered to be the golden baseline.展开更多
With the advancements in parameter-efficient transfer learning techniques,it has become feasible to leverage large pre-trained language models for downstream tasks under low-cost and low-resource conditions.However,ap...With the advancements in parameter-efficient transfer learning techniques,it has become feasible to leverage large pre-trained language models for downstream tasks under low-cost and low-resource conditions.However,applying this technique to multimodal knowledge transfer introduces a significant challenge:ensuring alignment across modalities while minimizing the number of additional parameters required for downstream task adaptation.This paper introduces UniTrans,a framework aimed at facilitating efficient knowledge transfer across multiple modalities.UniTrans leverages Vector-based Cross-modal Random Matrix Adaptation to enable fine-tuning with minimal parameter overhead.To further enhance modality alignment,we introduce two key components:the Multimodal Consistency Alignment Module and the Query-Augmentation Side Network,specifically optimized for scenarios with extremely limited trainable parameters.Extensive evaluations on various cross-modal downstream tasks demonstrate that our approach surpasses state-of-the-art methods while using just 5%of their trainable parameters.Additionally,it achieves superior performance compared to fully fine-tuned models on certain benchmarks.展开更多
We experimentally and numerically investigate CH_3I molecular alignment by using a femtosecond laser and a hexapole. The hexapole provides the single |111〉rotational state condition at 4.5-kV hexapole rod voltage. Ba...We experimentally and numerically investigate CH_3I molecular alignment by using a femtosecond laser and a hexapole. The hexapole provides the single |111〉rotational state condition at 4.5-kV hexapole rod voltage. Based on this single rotational state, an enhanced alignment degree of 0.73 is achieved. Our experimental results are in agreement with the simulation results. We experimentally obtain the ion velocity map images and show the influence of the initial rotational-state population. With the I+ion images and angular distributions at different pump-probe delay time, the alignment and anti-alignment phenomena are further demonstrated. The molecules will be under field-free conditions when the laser effect disappears completely at the full revival time. Our work shows that the quantum control and spatial control on CH_3I molecules can be realized and molecular coordinate frame can be obtained for further molecular experiment.展开更多
The alignment-dependent photoelectron spectrum is a valuable tool for mapping out the electronic structure of molecular orbitals.However,this approach may not be applicable to all molecules,such as CO_(2),as the ioniz...The alignment-dependent photoelectron spectrum is a valuable tool for mapping out the electronic structure of molecular orbitals.However,this approach may not be applicable to all molecules,such as CO_(2),as the ionization process in a linearly polarized laser field involves contributions from orbitals other than the highest occupied molecular orbital(HOMO).Here,we conducted a theoretical investigation into the ionization process of N_(2) and CO_(2) in near-circularly polarized laser field using the Coulomb-corrected strong-field approximation(CCSFA)method for molecules.In particular,we introduced a generalized dressed state into the CCSFA method in order to account for the impact of the laser field on the molecular initial state.The simulated alignment-dependent photoelectron momentum distribution(PMD)of the two molecules exhibited markedly disparate behaviors,which were in excellent agreement with the previous experimental observations reported in[Phys.Rev.A 102,013117(2020)].Our findings indicate that under a near-circularly polarized laser field,the alignment-dependent PMD of molecules is primarily sourced from the HOMO,in contrast to the situation under a linearly polarized laser field.Moreover,a satisfactory correlation between the alignment-dependent angular distribution and the orbital symmetry was observed,which suggests an effective approach for molecular orbital imaging.展开更多
The objective of this study is to address semantic misalignment and insufficient accuracy in edge detail and discrimination detection,which are common issues in deep learning-based change detection methods relying on ...The objective of this study is to address semantic misalignment and insufficient accuracy in edge detail and discrimination detection,which are common issues in deep learning-based change detection methods relying on encoding and decoding frameworks.In response to this,we propose a model called FlowDual-PixelClsObjectMec(FPCNet),which innovatively incorporates dual flow alignment technology in the decoding stage to rectify semantic discrepancies through streamlined feature correction fusion.Furthermore,the model employs an object-level similarity measurement coupled with pixel-level classification in the PixelClsObjectMec(PCOM)module during the final discrimination stage,significantly enhancing edge detail detection and overall accuracy.Experimental evaluations on the change detection dataset(CDD)and building CDD demonstrate superior performance,with F1 scores of 95.1%and 92.8%,respectively.Our findings indicate that the FPCNet outperforms the existing algorithms in stability,robustness,and other key metrics.展开更多
A novel algorithm of 3-D surface image registration is proposed. It makes use of the array information of 3-D points and takes vector/vertex-like features as the basis of the matching. That array information of 3-D po...A novel algorithm of 3-D surface image registration is proposed. It makes use of the array information of 3-D points and takes vector/vertex-like features as the basis of the matching. That array information of 3-D points can be easily obtained when capturing original 3-D images. The iterative least-mean-squared （LMS） algorithm is applied to optimizing adaptively the transformation matrix parameters. These can effectively improve the registration performance and hurry up the matching process. Experimental results show that it can reach a good subjective impression on aligned 3-D images. Although the algorithm focuses primarily on the human head model, it can also be used for other objects with small modifications.展开更多
Image/video stitching is a technology for solving the field of view(FOV)limitation of images/videos.It stitches multiple overlapping images/videos to generate a wide-FOV image/video,and has been used in various fields...Image/video stitching is a technology for solving the field of view(FOV)limitation of images/videos.It stitches multiple overlapping images/videos to generate a wide-FOV image/video,and has been used in various fields such as sports broadcasting,video surveillance,street view,and entertainment.This survey reviews image/video stitching algorithms,with a particular focus on those developed in recent years.Image stitching first calculates the corresponding relationships between multiple overlapping images,deforms and aligns the matched images,and then blends the aligned images to generate a wide-FOV image.A seamless method is always adopted to eliminate such potential flaws as ghosting and blurring caused by parallax or objects moving across the overlapping regions.Video stitching is the further extension of image stitching.It usually stitches selected frames of original videos to generate a stitching template by performing image stitching algorithms,and the subsequent frames can then be stitched according to the template.Video stitching is more complicated with moving objects or violent camera movement,because these factors introduce jitter,shakiness,ghosting,and blurring.Foreground detection technique is usually combined into stitching to eliminate ghosting and blurring,while video stabilization algorithms are adopted to solve the jitter and shakiness.This paper further discusses panoramic stitching as a special-extension of image/video stitching.Panoramic stitching is currently the most widely used application in stitching.This survey reviews the latest image/video stitching methods,and introduces the fundamental principles/advantages/weaknesses of image/video stitching algorithms.Image/video stitching faces long-term challenges such as wide baseline,large parallax,and low-texture problem in the overlapping region.New technologies may present new opportunities to address these issues,such as deep learning-based semantic correspondence,and 3D image stitching.Finally,this survey discusses the challenges of image/video stitching and proposes potential solutions.展开更多
Purpose: We investigated the margin recipes with different alignment techniques in the image-guided intensity-modulated radiotherapy (IMRT) of whole pelvis prostate cancer patients. Materials and Methods: Forty-eight ...Purpose: We investigated the margin recipes with different alignment techniques in the image-guided intensity-modulated radiotherapy (IMRT) of whole pelvis prostate cancer patients. Materials and Methods: Forty-eight computed tomography (CT) scans of eight prostate cancer patients were investigated. Each patient had an initial planning CT scan and 5 consecutive serial CT scans during the course of treatment, all of which were acquired using 3 mm slice separation and 0.94 mm resolution in the axial plane at 120 kVp, on a PQ 5000 CT scanner. Three different whole pelvis planning margin recipes, ranging from 3 to 13 mm, were investigated. A unique IMRT plan was created with each PTV on the initial CT scan, and was then registered to the 5 serial CT scans, by bony alignment or by prostate gland-based alignment. The dose computed on each serial CT scans was accumulated back to the initial CT scan using deformable image registration for final dosimetric evaluation of the interplay of the margin selection and alignment methods. Results: Bony alignment and prostate gland-based alignment gave very similar result to the pelvic lymphatic nodes (PLNs), regardless of its margin around. The prostate gland-based alignment greatly enhanced the coverage to the prostate and SV, especially with small margins. Meanwhile, the soft-tissue alignment also raised the incidental dose to the rectum and reduces the dose to the bladder. With small to intermediate margins, only soft-tissue alignment gave acceptable mean coverage to SV. Margin of 13mm or more was needed for PLNs to maintain good target coverage. Conclusion: We commend prostate-based alignment along with margins less than or equal to 5mm around prostate and SV, and margins greater than or equal to 13 mm around the vascular spaces.展开更多
The scope of this project was to investigate the possibility of application of Image Processing Technique in the field of Shaft Alignment process. Misalignment of shaft using image processing software Visionbuilder wa...The scope of this project was to investigate the possibility of application of Image Processing Technique in the field of Shaft Alignment process. Misalignment of shaft using image processing software Visionbuilder was calculated. The further purpose of this project was to check whether the image processing technique can be used in bone transplant surgery. The model of the hip was used for the experimentation purpose. Image processing software Visionbuilder was used to match the profiles of the bone before implant and bone after implant.展开更多
Phase transition temperatures of p,n-alkyloxy benzoic acids (nOBA, n = 3 to 10 and 12) are investigated basing on the textural image analysis of liquid crystal. The analysis is carried out by the computation of Legend...Phase transition temperatures of p,n-alkyloxy benzoic acids (nOBA, n = 3 to 10 and 12) are investigated basing on the textural image analysis of liquid crystal. The analysis is carried out by the computation of Legendre moments. Textures of the homeotropically aligned compounds are recorded as a function of temperature using POM in arthroscopic mode attached to the hot stage and high resolution camera. A recurrence formula is used to compute the liquid crystal textures based on Legendre polynomial. The discontinuities and fluctuations in the values of Legendre moments as a function of temperature are related to the phase transition temperatures of the sample. This method is successful in conforming or detecting the phase transition temperatures and the present findings are comparable with literature.展开更多
While photo-thermoelectric(PTE)sensors and their ultrabroadband monitoring facilitate nondestructive testing,their conventional fabrication is insufficient for high-yield integration.Specifically,PTE devices faced cha...While photo-thermoelectric(PTE)sensors and their ultrabroadband monitoring facilitate nondestructive testing,their conventional fabrication is insufficient for high-yield integration.Specifically,PTE devices faced challenges in their crucial spatial-misalignment for separate fabrication processes per constituent.Herein,this work demonstrates mechanically alignable and all-dispenser-printable integration of carbon nanotube(CNT)functional PTE sensor devices by designing them with solutionprocessable ink-materials.This technique first accurately prints CNT channels,essential in PTE conversion,using higher-concentration inks,and integrates remaining constituents(dopants and conductive pastes)into single device structures at high-yield.This work further demonstrates that employing higher-concentration CNT inks,suitable for mechanical channel printing,also designs sensitive PTE sensors.These sensors serve stably as integrated devices on diverse functional substrates,facilitating ubiquitous non-destructive monitoring depending on features.Therefore,this work designs such CNT PTE integrated devices and the associated functional inspection appropriately for structures,sizes,and external environments(e.g.,temperature and humidity)of monitoring targets.展开更多
Enhancement of the SNR (signal to noise ratio) in single-molecule imaging is significantly important for improving image resolu-tion and distinguishing the fine structures of single molecules at a higher precision lev...Enhancement of the SNR (signal to noise ratio) in single-molecule imaging is significantly important for improving image resolu-tion and distinguishing the fine structures of single molecules at a higher precision level.Image processing techniques have dem-onstrated the remarkable capability to improve the SNR and the resolution level by breaking through some inherent limitations unresolved by instrument hardware optimization.In this paper, we focus on single-biomolecule imaging using atomic force mi-croscopy (AFM), a unique method in separated single-biomolecule imaging, and there was few suitable image processing tech-niques reported to increase the SNR of a single molecule’s AFM image.With the typical samples of separately dispersed DNA molecules, we replaced the classified averaging method, which is usually used when the molecules’ structure can be easily and repeatedly prepared, with the time averaging method to improve the SNR in a single molecule’s AFM imaging.Combining the time averaging technique with the image alignment method for the series of time-lapse AFM images of a single DNA molecule, we were able to improve the image’s SNR and recover some buried signals from the background noises.Furthermore, the fine structures of single molecules can potentially be further resolved if other image processing techniques are applied collaboratively with some newly developed imaging methods like Vibrating Mode Scanning Polarization Force Microscopy (VSPFM), and such combination will also be advantageous for the manipulation of single-biomolecules by AFM.In addition, the proposed algorithms for the estimations of drift, distortion and SNR are applicable for quantitative status characterization of AFM imaging.展开更多
Beam alignment depends on CCD real-time image analysis and processing.In order to improve the quality of the alignment,multiple filters are used in far-field and near-field image processings.These multiple filters are...Beam alignment depends on CCD real-time image analysis and processing.In order to improve the quality of the alignment,multiple filters are used in far-field and near-field image processings.These multiple filters are constituted of an average filter and a median filter in different connection sequences,so that they can deal with different kinds of noise.To reduce the effect of the unknown nonlinear relationship between motor running steps and deviation pixels,a feasible methodology is offered to improve this phenomenon and a fuzzy algorithm is applied to the motor feedback control process.Because of the fuzzy control it is not necessary to establish an accurate mathematical model,so the impact of the nonlinear relationship will be reduced.展开更多
X-ray computed tomography at the nanometer scale （nano-CT） offers a wide range of applications in scientific and industrial areas. Here we describe a reliable, user-friendly, and fast software package based on LabVI...X-ray computed tomography at the nanometer scale （nano-CT） offers a wide range of applications in scientific and industrial areas. Here we describe a reliable, user-friendly, and fast software package based on LabVIEW that may allow us to perform all procedures after the acquisition of raw projection images in order to obtain the inner structure of the investigated sample. A suitable image alignment process to address misalignment problems among image series due to mechanical manufacturing errors, thermal expansion, and other external factors has been considered, together with a novel fast parallel beam 3D reconstruction procedure that was developed ad hoc to perform the tomographic reconstruction. We have obtained remarkably improved reconstruction results at the Beijing Synchrotron Radiation Facility after the image calibration, the fundamental role of this image alignment procedure was confirmed, which minimizes the unwanted blurs and additional streaking artifacts that are always present in reconstructed slices. Moreover, this nano-CT image alignment and its associated 3D reconstruction procedure are fully based on LabVIEW routines, significantly reducing the data post-processing cycle, thus making the activity of the users faster and easier during experimental runs.展开更多
The problem of robust alignment of batches of images can be formulated as a low-rank matrix optimization problem, relying on the similarity of well-aligned images. Going further, observing that the images to be aligne...The problem of robust alignment of batches of images can be formulated as a low-rank matrix optimization problem, relying on the similarity of well-aligned images. Going further, observing that the images to be aligned are sampled from a union of low-rank subspaces, we propose a new method based on subspace recovery techniques to provide more robust and accurate alignment. The proposed method seeks a set of domain transformations which are applied to the unaligned images so that the resulting images are made as similar as possible. The resulting optimization problem can be linearized as a series of convex optimization problems which can be solved by alternative sparsity pursuit techniques. Compared to existing methods like robust alignment by sparse and low-rank models, the proposed method can more effectively solve the batch image alignment problem,and extract more similar structures from the misaligned images.展开更多
文摘This paper presents a corner-based image alignment algorithm based on the procedures of corner-based template matching and geometric parameter estimation. This algorithm consists of two stages: 1) training phase, and 2) matching phase. In the training phase, a corner detection algorithm is used to extract the corners. These corners are then used to build the pyramid images. In the matching phase, the corners are obtained using the same corner detection algorithm. The similarity measure is then determined by the differences of gradient vector between the corners obtained in the template image and the inspection image, respectively. A parabolic function is further applied to evaluate the geometric relationship between the template and the inspection images. Results show that the corner-based template matching outperforms the original edge-based template matching in efficiency, and both of them are robust against non-liner light changes. The accuracy and precision of the corner-based image alignment are competitive to that of edge-based image alignment under the same environment. In practice, the proposed algorithm demonstrates its precision, efficiency and robustness in image alignment for real world applications.
基金supported by the Fund of Key Laboratory of Biomedical Engineering of Hainan Province(No.BME20240001)the STI2030-Major Projects(No.2021ZD0200104)the National Natural Science Foundations of China under Grant 61771437.
文摘Deep learning networks are increasingly exploited in the field of neuronal soma segmentation.However,annotating dataset is also an expensive and time-consuming task.Unsupervised domain adaptation is an effective method to mitigate the problem,which is able to learn an adaptive segmentation model by transferring knowledge from a rich-labeled source domain.In this paper,we propose a multi-level distribution alignment-based unsupervised domain adaptation network(MDA-Net)for segmentation of 3D neuronal soma images.Distribution alignment is performed in both feature space and output space.In the feature space,features from different scales are adaptively fused to enhance the feature extraction capability for small target somata and con-strained to be domain invariant by adversarial adaptation strategy.In the output space,local discrepancy maps that can reveal the spatial structures of somata are constructed on the predicted segmentation results.Then thedistribution alignment is performed on the local discrepancies maps across domains to obtain a superior discrepancy map in the target domain,achieving refined segmentation performance of neuronal somata.Additionally,after a period of distribution align-ment procedure,a portion of target samples with high confident pseudo-labels are selected as training data,which assist in learning a more adaptive segmentation network.We verified the superiority of the proposed algorithm by comparing several domain adaptation networks on two 3D mouse brain neuronal somata datasets and one macaque brain neuronal soma dataset.
基金supported by the National Natural Science Foundation of China(Nos.U20A20171,61802347,61972347,and 61773348)the Science Foundation of Zhejiang Province(Nos.LY21F020027,LGF20H180002,and LSD19H180003)。
文摘Recently, the generative adversarial network(GAN) has been extensively applied to the cross-modality conversion of medical images and has shown outstanding performance than other image conversion algorithms. Hence, we propose a novel GAN-based multi-domain registration method named multiscale diffeomorphic jointed network of registration and synthesis(MDJRS-Net). The deviation of the generator of the GAN-based approach affects the alignment phase, so a joint training strategy is introduced to improve the performance of the generator, which feedbacks the structural loss contained in the deformation field. Meanwhile, the nature of diffeomorphism can enable the network to generate deformation fields with more anatomical properties. The average dice score(Dice) is improved by 1.95% for the computer tomography venous(CTV) to magnetic resonance imaging(MRI) registration task and by 1.92% for the CTV to computer tomography plain(CTP) task compared with the other methods.
文摘Global motion estimation (GME) algorithms are widely applied to computer vision and video processing. In the previous works, the image resolutions are usually low for the real-time requirement (e.g. video stabilization). However, in some mobile devices applications (e.g. image sequence panoramic stitching), the high resolution is necessary to obtain satisfactory quality of panoramic image. However, the computational cost will become too expensive to be suitable for the low power consumption requirement of mobile device. The full search algorithm can obtain the global minimum with extremely computational cost, while the typical fast algorithms may suffer from the local minimum problem. This paper proposed a fast algorithm to deal with 2560 × 1920 high-resolution (HR) image sequences. The proposed method estimates the motion vector by a two-level coarse-to-fine scheme which only exploits sparse reference blocks (25 blocks in this paper) in each level to determine the global motion vector, thus the computational costs are significantly decreased. In order to increase the effective search range and robustness, the predictive motion vector (PMV) technique is used in this work. By the comparisons of computational complexity, the proposed algorithm costs less addition operations than the typical Three-Step Search algorithm (TSS) for estimating the global motion of the HR images without the local minimum problem. The quantitative evaluations show that our method is comparable to the full search algorithm (FSA) which is considered to be the golden baseline.
文摘With the advancements in parameter-efficient transfer learning techniques,it has become feasible to leverage large pre-trained language models for downstream tasks under low-cost and low-resource conditions.However,applying this technique to multimodal knowledge transfer introduces a significant challenge:ensuring alignment across modalities while minimizing the number of additional parameters required for downstream task adaptation.This paper introduces UniTrans,a framework aimed at facilitating efficient knowledge transfer across multiple modalities.UniTrans leverages Vector-based Cross-modal Random Matrix Adaptation to enable fine-tuning with minimal parameter overhead.To further enhance modality alignment,we introduce two key components:the Multimodal Consistency Alignment Module and the Query-Augmentation Side Network,specifically optimized for scenarios with extremely limited trainable parameters.Extensive evaluations on various cross-modal downstream tasks demonstrate that our approach surpasses state-of-the-art methods while using just 5%of their trainable parameters.Additionally,it achieves superior performance compared to fully fine-tuned models on certain benchmarks.
基金Project supported by the National Natural Science Foundation of China(Grant Nos.11574116,11534004,10704028,and 11474123)the Natural Science Foundation of Jilin Province,China(Grant No.20170101154JC)
文摘We experimentally and numerically investigate CH_3I molecular alignment by using a femtosecond laser and a hexapole. The hexapole provides the single |111〉rotational state condition at 4.5-kV hexapole rod voltage. Based on this single rotational state, an enhanced alignment degree of 0.73 is achieved. Our experimental results are in agreement with the simulation results. We experimentally obtain the ion velocity map images and show the influence of the initial rotational-state population. With the I+ion images and angular distributions at different pump-probe delay time, the alignment and anti-alignment phenomena are further demonstrated. The molecules will be under field-free conditions when the laser effect disappears completely at the full revival time. Our work shows that the quantum control and spatial control on CH_3I molecules can be realized and molecular coordinate frame can be obtained for further molecular experiment.
基金supported by the National Natural Science Foundation of China(Grant No.12274273).
文摘The alignment-dependent photoelectron spectrum is a valuable tool for mapping out the electronic structure of molecular orbitals.However,this approach may not be applicable to all molecules,such as CO_(2),as the ionization process in a linearly polarized laser field involves contributions from orbitals other than the highest occupied molecular orbital(HOMO).Here,we conducted a theoretical investigation into the ionization process of N_(2) and CO_(2) in near-circularly polarized laser field using the Coulomb-corrected strong-field approximation(CCSFA)method for molecules.In particular,we introduced a generalized dressed state into the CCSFA method in order to account for the impact of the laser field on the molecular initial state.The simulated alignment-dependent photoelectron momentum distribution(PMD)of the two molecules exhibited markedly disparate behaviors,which were in excellent agreement with the previous experimental observations reported in[Phys.Rev.A 102,013117(2020)].Our findings indicate that under a near-circularly polarized laser field,the alignment-dependent PMD of molecules is primarily sourced from the HOMO,in contrast to the situation under a linearly polarized laser field.Moreover,a satisfactory correlation between the alignment-dependent angular distribution and the orbital symmetry was observed,which suggests an effective approach for molecular orbital imaging.
文摘The objective of this study is to address semantic misalignment and insufficient accuracy in edge detail and discrimination detection,which are common issues in deep learning-based change detection methods relying on encoding and decoding frameworks.In response to this,we propose a model called FlowDual-PixelClsObjectMec(FPCNet),which innovatively incorporates dual flow alignment technology in the decoding stage to rectify semantic discrepancies through streamlined feature correction fusion.Furthermore,the model employs an object-level similarity measurement coupled with pixel-level classification in the PixelClsObjectMec(PCOM)module during the final discrimination stage,significantly enhancing edge detail detection and overall accuracy.Experimental evaluations on the change detection dataset(CDD)and building CDD demonstrate superior performance,with F1 scores of 95.1%and 92.8%,respectively.Our findings indicate that the FPCNet outperforms the existing algorithms in stability,robustness,and other key metrics.
文摘A novel algorithm of 3-D surface image registration is proposed. It makes use of the array information of 3-D points and takes vector/vertex-like features as the basis of the matching. That array information of 3-D points can be easily obtained when capturing original 3-D images. The iterative least-mean-squared （LMS） algorithm is applied to optimizing adaptively the transformation matrix parameters. These can effectively improve the registration performance and hurry up the matching process. Experimental results show that it can reach a good subjective impression on aligned 3-D images. Although the algorithm focuses primarily on the human head model, it can also be used for other objects with small modifications.
基金the National Natural Science Foundation of China(61872023).
文摘Image/video stitching is a technology for solving the field of view(FOV)limitation of images/videos.It stitches multiple overlapping images/videos to generate a wide-FOV image/video,and has been used in various fields such as sports broadcasting,video surveillance,street view,and entertainment.This survey reviews image/video stitching algorithms,with a particular focus on those developed in recent years.Image stitching first calculates the corresponding relationships between multiple overlapping images,deforms and aligns the matched images,and then blends the aligned images to generate a wide-FOV image.A seamless method is always adopted to eliminate such potential flaws as ghosting and blurring caused by parallax or objects moving across the overlapping regions.Video stitching is the further extension of image stitching.It usually stitches selected frames of original videos to generate a stitching template by performing image stitching algorithms,and the subsequent frames can then be stitched according to the template.Video stitching is more complicated with moving objects or violent camera movement,because these factors introduce jitter,shakiness,ghosting,and blurring.Foreground detection technique is usually combined into stitching to eliminate ghosting and blurring,while video stabilization algorithms are adopted to solve the jitter and shakiness.This paper further discusses panoramic stitching as a special-extension of image/video stitching.Panoramic stitching is currently the most widely used application in stitching.This survey reviews the latest image/video stitching methods,and introduces the fundamental principles/advantages/weaknesses of image/video stitching algorithms.Image/video stitching faces long-term challenges such as wide baseline,large parallax,and low-texture problem in the overlapping region.New technologies may present new opportunities to address these issues,such as deep learning-based semantic correspondence,and 3D image stitching.Finally,this survey discusses the challenges of image/video stitching and proposes potential solutions.
文摘Purpose: We investigated the margin recipes with different alignment techniques in the image-guided intensity-modulated radiotherapy (IMRT) of whole pelvis prostate cancer patients. Materials and Methods: Forty-eight computed tomography (CT) scans of eight prostate cancer patients were investigated. Each patient had an initial planning CT scan and 5 consecutive serial CT scans during the course of treatment, all of which were acquired using 3 mm slice separation and 0.94 mm resolution in the axial plane at 120 kVp, on a PQ 5000 CT scanner. Three different whole pelvis planning margin recipes, ranging from 3 to 13 mm, were investigated. A unique IMRT plan was created with each PTV on the initial CT scan, and was then registered to the 5 serial CT scans, by bony alignment or by prostate gland-based alignment. The dose computed on each serial CT scans was accumulated back to the initial CT scan using deformable image registration for final dosimetric evaluation of the interplay of the margin selection and alignment methods. Results: Bony alignment and prostate gland-based alignment gave very similar result to the pelvic lymphatic nodes (PLNs), regardless of its margin around. The prostate gland-based alignment greatly enhanced the coverage to the prostate and SV, especially with small margins. Meanwhile, the soft-tissue alignment also raised the incidental dose to the rectum and reduces the dose to the bladder. With small to intermediate margins, only soft-tissue alignment gave acceptable mean coverage to SV. Margin of 13mm or more was needed for PLNs to maintain good target coverage. Conclusion: We commend prostate-based alignment along with margins less than or equal to 5mm around prostate and SV, and margins greater than or equal to 13 mm around the vascular spaces.
文摘The scope of this project was to investigate the possibility of application of Image Processing Technique in the field of Shaft Alignment process. Misalignment of shaft using image processing software Visionbuilder was calculated. The further purpose of this project was to check whether the image processing technique can be used in bone transplant surgery. The model of the hip was used for the experimentation purpose. Image processing software Visionbuilder was used to match the profiles of the bone before implant and bone after implant.
文摘Phase transition temperatures of p,n-alkyloxy benzoic acids (nOBA, n = 3 to 10 and 12) are investigated basing on the textural image analysis of liquid crystal. The analysis is carried out by the computation of Legendre moments. Textures of the homeotropically aligned compounds are recorded as a function of temperature using POM in arthroscopic mode attached to the hot stage and high resolution camera. A recurrence formula is used to compute the liquid crystal textures based on Legendre polynomial. The discontinuities and fluctuations in the values of Legendre moments as a function of temperature are related to the phase transition temperatures of the sample. This method is successful in conforming or detecting the phase transition temperatures and the present findings are comparable with literature.
基金supported by a part of ACT-X(JPMJAX23KL)and Mirai Programme(JPMJMI23G1):JST,KAKENHI(JP22H01553,JP22H01555,JP23H00169,JP23K19125,JP24K01288,and JP24K17325):JSPS,and grants(Murata Science Foundation,Matsuo Foundation,Sumitomo Electric Groups CSR Foundation,Takano Science Foundation,Thermal&Electric Energy Technology Foundation,Precise Measurement Technology Promotion Foundation,Suzuki Foundation,Iwatani Naoji Foundation,Futaba Foundation,Konica Minolta Science and Technology Foundation,Fuji Seal Foundation,Telecommunications Advancement Foundation,TEPCO Memorial Foundation,Paloma CSR Foundation,Takano Life Science Research Foundation,The Foundation for The Promotion of Ion Engineering,Hattori Hokokai Foundation,Mechanical Social Systems Foundation,Kayamori Foundation of Informational Science Advancement,Shimadzu Science Foundation,Japan Power Academy,Tokuyama Science Foundation,Yashima Environment Technology Foundation,Tateisi Science and Technology Foundation,The Kajima Foundation,Japan Keirin Autorace Foundation,Amano Institute of Technology,and Kanagawa Institute of Industrial Science and Technology).
文摘While photo-thermoelectric(PTE)sensors and their ultrabroadband monitoring facilitate nondestructive testing,their conventional fabrication is insufficient for high-yield integration.Specifically,PTE devices faced challenges in their crucial spatial-misalignment for separate fabrication processes per constituent.Herein,this work demonstrates mechanically alignable and all-dispenser-printable integration of carbon nanotube(CNT)functional PTE sensor devices by designing them with solutionprocessable ink-materials.This technique first accurately prints CNT channels,essential in PTE conversion,using higher-concentration inks,and integrates remaining constituents(dopants and conductive pastes)into single device structures at high-yield.This work further demonstrates that employing higher-concentration CNT inks,suitable for mechanical channel printing,also designs sensitive PTE sensors.These sensors serve stably as integrated devices on diverse functional substrates,facilitating ubiquitous non-destructive monitoring depending on features.Therefore,this work designs such CNT PTE integrated devices and the associated functional inspection appropriately for structures,sizes,and external environments(e.g.,temperature and humidity)of monitoring targets.
基金supported by the National Basic Research Program of China (2006CB932505 and 2007CB936004)
文摘Enhancement of the SNR (signal to noise ratio) in single-molecule imaging is significantly important for improving image resolu-tion and distinguishing the fine structures of single molecules at a higher precision level.Image processing techniques have dem-onstrated the remarkable capability to improve the SNR and the resolution level by breaking through some inherent limitations unresolved by instrument hardware optimization.In this paper, we focus on single-biomolecule imaging using atomic force mi-croscopy (AFM), a unique method in separated single-biomolecule imaging, and there was few suitable image processing tech-niques reported to increase the SNR of a single molecule’s AFM image.With the typical samples of separately dispersed DNA molecules, we replaced the classified averaging method, which is usually used when the molecules’ structure can be easily and repeatedly prepared, with the time averaging method to improve the SNR in a single molecule’s AFM imaging.Combining the time averaging technique with the image alignment method for the series of time-lapse AFM images of a single DNA molecule, we were able to improve the image’s SNR and recover some buried signals from the background noises.Furthermore, the fine structures of single molecules can potentially be further resolved if other image processing techniques are applied collaboratively with some newly developed imaging methods like Vibrating Mode Scanning Polarization Force Microscopy (VSPFM), and such combination will also be advantageous for the manipulation of single-biomolecules by AFM.In addition, the proposed algorithms for the estimations of drift, distortion and SNR are applicable for quantitative status characterization of AFM imaging.
基金supported by grants from the Chinese and Israeli cooperation project on high power laser technology(2010DFB70490)
文摘Beam alignment depends on CCD real-time image analysis and processing.In order to improve the quality of the alignment,multiple filters are used in far-field and near-field image processings.These multiple filters are constituted of an average filter and a median filter in different connection sequences,so that they can deal with different kinds of noise.To reduce the effect of the unknown nonlinear relationship between motor running steps and deviation pixels,a feasible methodology is offered to improve this phenomenon and a fuzzy algorithm is applied to the motor feedback control process.Because of the fuzzy control it is not necessary to establish an accurate mathematical model,so the impact of the nonlinear relationship will be reduced.
基金Supported by Major State Basic Research Development Program(2012CB825800)Science Fund for Creative Research Groups(11321503)+2 种基金Knowledge Innovation Program of The Chinese Academy of Sciences(KJCX2-YW-N42)National Natural Science Foundation of China(11179004,10979055,11205189,11205157)Fundamental Research Funds for the Central Universities(WK2310000021)
文摘X-ray computed tomography at the nanometer scale （nano-CT） offers a wide range of applications in scientific and industrial areas. Here we describe a reliable, user-friendly, and fast software package based on LabVIEW that may allow us to perform all procedures after the acquisition of raw projection images in order to obtain the inner structure of the investigated sample. A suitable image alignment process to address misalignment problems among image series due to mechanical manufacturing errors, thermal expansion, and other external factors has been considered, together with a novel fast parallel beam 3D reconstruction procedure that was developed ad hoc to perform the tomographic reconstruction. We have obtained remarkably improved reconstruction results at the Beijing Synchrotron Radiation Facility after the image calibration, the fundamental role of this image alignment procedure was confirmed, which minimizes the unwanted blurs and additional streaking artifacts that are always present in reconstructed slices. Moreover, this nano-CT image alignment and its associated 3D reconstruction procedure are fully based on LabVIEW routines, significantly reducing the data post-processing cycle, thus making the activity of the users faster and easier during experimental runs.
基金supported by the National Natural Science Foundation of China (Grant Nos. 61573150, 61573152, 61370185, 61403085, and 51275094)Guangzhou Project Nos. 201604016113 and 201604046018
文摘The problem of robust alignment of batches of images can be formulated as a low-rank matrix optimization problem, relying on the similarity of well-aligned images. Going further, observing that the images to be aligned are sampled from a union of low-rank subspaces, we propose a new method based on subspace recovery techniques to provide more robust and accurate alignment. The proposed method seeks a set of domain transformations which are applied to the unaligned images so that the resulting images are made as similar as possible. The resulting optimization problem can be linearized as a series of convex optimization problems which can be solved by alternative sparsity pursuit techniques. Compared to existing methods like robust alignment by sparse and low-rank models, the proposed method can more effectively solve the batch image alignment problem,and extract more similar structures from the misaligned images.
