Dual-convolutional neural network-enhanced strain estimation method for optical coherence elastography.

Strain estimation is vital in phase-sensitive optical coherence elastography (PhS-OCE). In this Letter, we introduce a novel, to the best of our knowledge, method to improve strain estimation by using a dual-convolutional neural network (Dual-CNN). This approach requires two sets of PhS-OCE systems: a high-resolution system for high-quality training data and a cost-effective standard-resolution system for practical measurements. During training, high-resolution strain results acquired from the former system and the pre-existing strain estimation CNN serve as label data, while the narrowed light source-acquired standard-resolution phase results act as input data. By training a new network with this data, high-quality strain results can be estimated from standard-resolution PhS-OCE phase results. Comparison experiments show that the proposed Dual-CNN can preserve the strain quality even when the light source bandwidth is reduced by over 80%.

Deep learning-based method for non-uniform motion-induced error reduction in dynamic microscopic 3D shape measurement.

The non-uniform motion-induced error reduction in dynamic fringe projection profilometry is complex and challenging. Recently, deep learning (DL) has been successfully applied to many complex optical problems with strong nonlinearity and exhibits excellent performance. Inspired by this, a deep learning-based method is developed for non-uniform motion-induced error reduction by taking advantage of the powerful ability of nonlinear fitting. First, a specially designed dataset of motion-induced error reduction is generated for network training by incorporating complex nonlinearity. Then, the corresponding DL-based architecture is proposed and it contains two parts: in the first part, a fringe compensation module is developed as network pre-processing to reduce the phase error caused by fringe discontinuity; in the second part, a deep neural network is employed to extract the high-level features of error distribution and establish a pixel-wise hidden nonlinear mapping between the phase with motion-induced error and the ideal one. Both simulations and real experiments demonstrate the feasibility of the proposed method in dynamic macroscopic measurement.

Generic saturation-induced phase error correction for structured light 3D shape measurement.

Intensity saturation is a challenging problem in structured light 3D shape measurement. Most of the existing methods achieve high dynamic range (HDR) measurement by sacrificing measurement speed, making them limited in high-speed dynamic applications. This Letter proposes a generic efficient saturation-induced phase error correction method for HDR measurement without increasing any fringe patterns. We first theoretically analyze the saturated signal model and deduce the periodic characteristic of saturation-induced phase error. Based on this, we specially design a saturation-induced phase error correction method by joint Fourier analysis and Hilbert transform. Furthermore, the relationship among phase error, saturation degree, and number of phase-shifting steps is established by numerical simulation. Since the proposed method requires no extra captured images or complicated intensity calibration, it is extremely convenient in implementation and is applicable to performing high-speed 3D shape measurements. Simulations and experiments verify the feasibility of the proposed method.

Super-resolution reconstruction of speckle phase in depth-resolved wavelength scanning interference using the total least-squares analysis.

Depth-resolved wavelength scanning interferometry (DRWSI) is a tomographic imaging tool that employs phase measurement to visualize micro-displacement inside a sample. It is well known that the depth resolution of DRWSI is restricted by a wavelength scanning range. Recently, a nonlinear least-squares analysis (NLS) algorithm was proposed to overcome the limitation of the wavelength scanning range to achieve super-resolution; however, the NLS failed to measure speckle surfaces owing to the sensibility of initial values. To the best of our knowledge, the improvement of depth resolution on measuring a speckle surface remains an open issue for DRWSI. For this study, we redesigned the signal processing algorithm for DRWSI to refine the depth resolution when considering the case of speckle phase measurement. It is mathematically shown that the DRWSI's signal is derived as a model of total least-squares analysis (TLSA). Subsequently, a super-resolution of the speckle phase map was obtained using a singular value decomposition. Further, a numerical simulation to measure the micro-displacements for speckle surfaces was performed to validate the TLSA, and the results show that it can precisely reconstruct the displacements of layers whose depth distance is 5 µm. This study thus provides an opportunity to improve the DRWSI's depth resolution.

Wavenumber synthesis approach to high-resolution wavenumber scanning interference using a mode-hoped laser.

A new method for the synthesis of wavenumber series before and after mode hopping is proposed for depth-resolved wavenumber scanning interferometry. The classical Fourier transform is not suitable for mode hopping; consequently, the wavenumber scanning range of diode lasers is rather narrow, reducing the depth resolution and measurement accuracy. We show that the discontinuity in wavenumber domain interferograms caused by mode hopping can be removed by introducing the phase compensation of the interference spectrum. Thus, the wavenumber series before and after mode hopping can be synthesized. Experiments and numerical simulations validate the proposed method, and the measurement error is within 5nm.

Highly sensitive, wide dynamic range displacement sensor combining chromatic confocal system and phase-sensitive spectral optical coherence tomography.

A displacement sensor with nanometer-sensitivity and a submillimeter dynamic range is proposed. It integrates a chromatic confocal system and phase-sensitive spectral optical coherence tomography (PhS-SOCT) into the fiber-based Michelson interferometer and codes interference and confocal signals with spectral multiplexing. A displacement is evaluated using depth-resolved phase information decoded from the interference signal, which is unwrapped based on the position information decoded from the confocal signal. A sensor system with a 0.102mm dynamic range was built to validate the method. The temperature induced sample surface displacement was measured with a root mean square error of 3.9nm.

Improvement of depth resolution in depth-resolved wavenumber-scanning interferometry using wavenumber-domain least-squares algorithm: comparison and experiment.

It is important to improve the depth resolution in depth-resolved wavenumber-scanning interferometry (DRWSI) owing to the limited range of wavenumber scanning. In this work, a new nonlinear iterative least-squares algorithm called the wavenumber-domain least-squares algorithm (WLSA) is proposed for evaluating the phase of DRWSI. The simulated and experimental results of the Fourier transform (FT), complex-number least-squares algorithm (CNLSA), eigenvalue-decomposition and least-squares algorithm (EDLSA), and WLSA were compared and analyzed. According to the results, the WLSA is less dependent on the initial values, and the depth resolution δz is approximately changed from δz to δz/6. Thus, the WLSA exhibits a better performance than the FT, CNLSA, and EDLSA.

Multimodal Fusion Network for 3-D Lane Detection.

3-D lane detection is a challenging task due to the diversity of lanes, occlusion, dazzle light, and so on. Traditional methods usually use highly specialized handcrafted features and carefully designed postprocessing to detect them. However, these methods are based on strong assumptions and single modal so that they are easily scalable and have poor performance. In this article, a multimodal fusion network (MFNet) is proposed through using multihead nonlocal attention and feature pyramid for 3-D lane detection. It includes three parts: multihead deformable transformation (MDT) module, multidirectional attention feature pyramid fusion (MA-FPF) module, and top-view lane prediction (TLP) ones. First, MDT is presented to learn and mine multimodal features from RGB images, depth maps, and point cloud data (PCD) for achieving optimal lane feature extraction. Then, MA-FPF is designed to fuse multiscale features for presenting the vanish of lane features as the network deepens. Finally, TLP is developed to estimate 3-D lanes and predict their position. Experimental results on the 3-D lane synthetic and ONCE-3DLanes datasets demonstrate that the performance of the proposed MFNet outperforms the state-of-the-art methods in both qualitative and quantitative analyses and visual comparisons.

CLAD-Net: cross-layer aggregation attention network for real-time endoscopic instrument detection.

As medical treatments continue to advance rapidly, minimally invasive surgery (MIS) has found extensive applications across various clinical procedures. Accurate identification of medical instruments plays a vital role in comprehending surgical situations and facilitating endoscopic image-guided surgical procedures. However, the endoscopic instrument detection poses a great challenge owing to the narrow operating space, with various interfering factors (e.g. smoke, blood, body fluids) and inevitable issues (e.g. mirror reflection, visual obstruction, illumination variation) in the surgery. To promote surgical efficiency and safety in MIS, this paper proposes a cross-layer aggregated attention detection network (CLAD-Net) for accurate and real-time detection of endoscopic instruments in complex surgical scenarios. We propose a cross-layer aggregation attention module to enhance the fusion of features and raise the effectiveness of lateral propagation of feature information. We propose a composite attention mechanism (CAM) to extract contextual information at different scales and model the importance of each channel in the feature map, mitigate the information loss due to feature fusion, and effectively solve the problem of inconsistent target size and low contrast in complex contexts. Moreover, the proposed feature refinement module (RM) enhances the network's ability to extract target edge and detail information by adaptively adjusting the feature weights to fuse different layers of features. The performance of CLAD-Net was evaluated using a public laparoscopic dataset Cholec80 and another set of neuroendoscopic dataset from Sun Yat-sen University Cancer Center. From both datasets and comparisons, CLAD-Net achieves the AP0.5 of 98.9% and 98.6%, respectively, that is better than advanced detection networks. A video for the real-time detection is presented in the following link: https://github.com/A0268/video-demo.

An Adaptive Image Segmentation Network for Surface Defect Detection.

Surface defect detection plays an essential role in industry, and it is challenging due to the following problems: 1) the similarity between defect and nondefect texture is very high, which eventually leads to recognition or classification errors and 2) the size of defects is tiny, which are much more difficult to be detected than larger ones. To address such problems, this article proposes an adaptive image segmentation network (AIS-Net) for pixelwise segmentation of surface defects. It consists of three main parts: multishuffle-block dilated convolution (MSDC), dual attention context guidance (DACG), and adaptive category prediction (ACP) modules, where MSDC is designed to merge the multiscale defect features for avoiding the loss of tiny defect feature caused by model depth, DACG is designed to capture more contextual information from the defect feature map for locating defect regions and obtaining clear segmentation boundaries, and ACP is used to make classification and regression for predicting defect categories. Experimental results show that the proposed AIS-Net is superior to the state-of-the-art approaches on four actual surface defect datasets (NEU-DET: 98.38% ± 0.03%, DAGM: 99.25% ± 0.02%, Magnetic-tile: 98.73% ± 0.13%, and MVTec: 99.72% ± 0.02%).

AANet: Adaptive Attention Network for COVID-19 Detection From Chest X-Ray Images.

Accurate and rapid diagnosis of COVID-19 using chest X-ray (CXR) plays an important role in large-scale screening and epidemic prevention. Unfortunately, identifying COVID-19 from the CXR images is challenging as its radiographic features have a variety of complex appearances, such as widespread ground-glass opacities and diffuse reticular-nodular opacities. To solve this problem, we propose an adaptive attention network (AANet), which can adaptively extract the characteristic radiographic findings of COVID-19 from the infected regions with various scales and appearances. It contains two main components: an adaptive deformable ResNet and an attention-based encoder. First, the adaptive deformable ResNet, which adaptively adjusts the receptive fields to learn feature representations according to the shape and scale of infected regions, is designed to handle the diversity of COVID-19 radiographic features. Then, the attention-based encoder is developed to model nonlocal interactions by self-attention mechanism, which learns rich context information to detect the lesion regions with complex shapes. Extensive experiments on several public datasets show that the proposed AANet outperforms state-of-the-art methods.

Eliminating the Permutation Ambiguity of Convolutive Blind Source Separation by Using Coupled Frequency Bins.

Blind source separation (BSS) is a typical unsupervised learning method that extracts latent components from their observations. In the meanwhile, convolutive BSS (CBSS) is particularly challenging as the observations are the mixtures of latent components as well as their delayed versions. CBSS is usually solved in frequency domain since convolutive mixtures in time domain is just instantaneous mixtures in frequency domain, which allows to recover source frequency components independently of each frequency bin by running ordinary BSS, and then concatenate them to form the Fourier transformation of source signals. Because BSS has inherent permutation ambiguity, this category of CBSS methods suffers from a common drawback: it is very difficult to choose the frequency components belonging to a specific source as they are estimated from different frequency bins using BSS. This paper presents a tensor framework that can completely eliminate the permutation ambiguity. By combining each frequency bin with an anchor frequency bin that is chosen arbitrarily in advance, we establish a new virtual BSS model where the corresponding correlation matrices comply with a block tensor decomposition (BTD) model. The essential uniqueness of BTD and the sparse structure of coupled mixing parameters allow the estimation of the mixing matrices free of permutation ambiguity. Extensive simulation results confirmed that the proposed algorithm could achieve higher separation accuracy compared with the state-of-the-art methods.

Robust Latent Subspace Learning for Image Classification.

This paper proposes a novel method, called robust latent subspace learning (RLSL), for image classification. We formulate an RLSL problem as a joint optimization problem over both the latent SL and classification model parameter predication, which simultaneously minimizes: 1) the regression loss between the learned data representation and objective outputs and 2) the reconstruction error between the learned data representation and original inputs. The latent subspace can be used as a bridge that is expected to seamlessly connect the origin visual features and their class labels and hence improve the overall prediction performance. RLSL combines feature learning with classification so that the learned data representation in the latent subspace is more discriminative for classification. To learn a robust latent subspace, we use a sparse item to compensate error, which helps suppress the interference of noise via weakening its response during regression. An efficient optimization algorithm is designed to solve the proposed optimization problem. To validate the effectiveness of the proposed RLSL method, we conduct experiments on diverse databases and encouraging recognition results are achieved compared with many state-of-the-arts methods.

Rate of Convergence of the FOCUSS Algorithm.

Focal underdetermined system solver (FOCUSS) is a powerful method for basis selection and sparse representation, where it employs the [Formula: see text]-norm with p ∈ (0,2) to measure the sparsity of solutions. In this paper, we give a systematical analysis on the rate of convergence of the FOCUSS algorithm with respect to p ∈ (0,2) . We prove that the FOCUSS algorithm converges superlinearly for and linearly for usually, but may superlinearly in some very special scenarios. In addition, we verify its rates of convergence with respect to p by numerical experiments.

Convergence analysis of the FOCUSS algorithm.

Focal Underdetermined System Solver (FOCUSS) is a powerful and easy to implement tool for basis selection and inverse problems. One of the fundamental problems regarding this method is its convergence, which remains unsolved until now. We investigate the convergence of the FOCUSS algorithm in this paper. We first give a rigorous derivation for the FOCUSS algorithm by exploiting the auxiliary function. Following this, we further prove its convergence by stability analysis.

Validating rationale of group-level component analysis based on estimating number of sources in EEG through model order selection.

This study addresses how to validate the rationale of group component analysis (CA) for blind source separation through estimating the number of sources in each individual EEG dataset via model order selection. Control children, typically reading children with risk for reading disability (RD), and children with RD participated in the experiment. Passive oddball paradigm was used for eliciting mismatch negativity during EEG data collection. Data were cleaned by two digital filters with pass bands of 1-30 Hz and 1-15 Hz and a wavelet filter with the pass band narrower than 1-12 Hz. Three model order selection methods were used to estimate the number of sources in each filtered EEG dataset. Under the filter with the pass band of 1-30 Hz, the numbers of sources were very similar among different individual EEG datasets and the group ICA would be suggested; regarding the other two filters with much narrower pass bands, the numbers of sources were relatively diverse, and then, applying group ICA would not be appropriate. Hence, before group ICA is performed, its rationale can be logically validated by the estimated number of sources in EEG data through model order selection.

Symmetric nonnegative matrix factorization: algorithms and applications to probabilistic clustering.

Nonnegative matrix factorization (NMF) is an unsupervised learning method useful in various applications including image processing and semantic analysis of documents. This paper focuses on symmetric NMF (SNMF), which is a special case of NMF decomposition. Three parallel multiplicative update algorithms using level 3 basic linear algebra subprograms directly are developed for this problem. First, by minimizing the Euclidean distance, a multiplicative update algorithm is proposed, and its convergence under mild conditions is proved. Based on it, we further propose another two fast parallel methods: α-SNMF and ß -SNMF algorithms. All of them are easy to implement. These algorithms are applied to probabilistic clustering. We demonstrate their effectiveness for facial image clustering, document categorization, and pattern clustering in gene expression.

Minimum-volume-constrained nonnegative matrix factorization: enhanced ability of learning parts.

Nonnegative matrix factorization (NMF) with minimum-volume-constraint (MVC) is exploited in this paper. Our results show that MVC can actually improve the sparseness of the results of NMF. This sparseness is L(0)-norm oriented and can give desirable results even in very weak sparseness situations, thereby leading to the significantly enhanced ability of learning parts of NMF. The close relation between NMF, sparse NMF, and the MVC_NMF is discussed first. Then two algorithms are proposed to solve the MVC_NMF model. One is called quadratic programming_MVC_NMF (QP_MVC_NMF) which is based on quadratic programming and the other is called negative glow_MVC_NMF (NG_MVC_NMF) because it uses multiplicative updates incorporating natural gradient ingeniously. The QP_MVC_NMF algorithm is quite efficient for small-scale problems and the NG_MVC_NMF algorithm is more suitable for large-scale problems. Simulations show the efficiency and validity of the proposed methods in applications of blind source separation and human face images analysis.

Detecting the number of clusters in n-way probabilistic clustering.

Recently, there has been a growing interest in multiway probabilistic clustering. Some efficient algorithms have been developed for this problem. However, not much attention has been paid on how to detect the number of clusters for the general n-way clustering (n ≥ 2). To fill this gap, this problem is investigated based on n-way algebraic theory in this paper. A simple, yet efficient, detection method is proposed by eigenvalue decomposition (EVD), which is easy to implement. We justify this method. In addition, its effectiveness is demonstrated by the experiments on both simulated and real-world data sets.

A note on Lewicki-Sejnowski gradient for learning overcomplete representations.

Overcomplete representations have greater robustness in noise environment and also have greater flexibility in matching structure in the data. Lewicki and Sejnowski (2000) proposed an efficient extended natural gradient for learning the overcomplete basis and developed an overcomplete representation approach. However, they derived their gradient by many approximations, and their proof is very complicated. To give a stronger theoretical basis, we provide a brief and more rigorous mathematical proof for this gradient in this note. In addition, we propose a more robust constrained Lewicki-Sejnowski gradient.