Search

Purpose Electrocardiogram (ECG) is one of the most essential tools for detecting heart problems. Till today most of the ECG records are available in paper form. It can be challenging and time-consuming to manually assess the ECG paper records. Hence, automated diagnosis and analysis are possible if we digitize such paper ECG records. Methods The proposed work aims to convert ECG paper records into a 1-D signal and generate an accurate diagnosis of heart-related problems using deep learning. Camera-captured ECG images or scanned ECG paper records are used for the proposed work. Effective pre-processing techniques are used for the removal of shadow from the images. A deep learning model is used to get a threshold value that separates ECG signal from its background and after applying various image processing techniques threshold ECG image gets converted into digital ECG. These digitized 1-D ECG signals are then passed to another deep learning model for the automated diagnosis of heart diseases into different classes such as ST-segment elevation myocardial infarction (STEMI), Left Bundle Branch Block (LBBB), Right Bundle Branch Block (RBBB), and T-wave abnormality. Results The accuracy of deep learning-based binarization is 97%. Further deep learning-based diagnosis approach of such digitized paper ECG records was having an accuracy of 94.4%. Conclusions The digitized ECG signals can be useful to various research organizations because the trends in heart problems can be determined and diagnosed from preserved paper ECG records. This approach can be easily implemented in areas where such expertise is not available. Supplementary Information The online version contains supplementary material available at 10.1007/s40846-021-00632-0.

International audience; Texture characterization of photographic papers is likely to provide scholars with valuable information regarding artistic practices. Currently, texture assessment remains mostly based on visual and manual inspections, implying long repetitive tasks prone to inter- and even intra-observer variability. Automated texture characterization and classification procedures are thus important tasks in historical studies of large databases of photographic papers, likely to provide quantitative and reproducible assessments of texture matches. Such procedures may, for instance, produce vital information on photographic prints of uncertain origins. The hyperbolic wavelet transform, because it relies on the use of different dilation factor along the horizontal and vertical axes, permits to construct robust and meaningful multiscale and anisotropic representation of textures. In the present contribution, we explore how unsupervised clustering strategies can be complemented both to assess the significance of extracted clusters and the strength of the contribution of each texture to its associated cluster. Graph based filterbank strategies are notably investigated with the aim to produce small size significant clusters. These tools are illustrated at work on a large database of about 2500 exposed and non exposed photographic papers carefully assembled and documented by the MoMA and P. Messier's foundation. Results are commented and interpreted.

International audience; Photographic paper texture characterization constitutes a challenging image processing task and an important stake both for manufacturers and art museums. The present contribution shows how the Hyperbolic Wavelet Transform, thanks to its joint multiscale and anisotropie nature, permits to achieve an accurate photographic paper texture analysis and characterization. A cepstral-type distance, constructed on the coefficients of the Hyperbolic Wavelet Transform, is then used to measure similarity between pairs of paper textures. Spectral clustering followed by Ascendant Hierarchical Clustering applied to the similarity matrix enables an unsupervised classification of photographic paper sheets. This methodology is applied to a test dataset made available in the framework of the Historic Photographic Paper Classification Challenge. The relevance of the proposed texture characterization and classification procedure is assessed by comparisons against the database documentation provided by experts.

International audience; This paper investigates the classification of photographic paper tex-tures using visual descriptors. Such classification is called fine grain due to the very low inter-class variability. We propose a novel image representation for photographic paper texture categorization, relying on the incorporation of a powerful local descriptor into an efficient higher-order model deviation where texture is represented by com-puting statistics on the occurrences of specific local visual patterns. We perform an evaluation on two different challenging datasets of photographic paper textures and show such advanced methods in-deed outperforms existing descriptors.

Feature extraction and representation plays a vital role in multimedia processing. It is still a challenge in computer vision system to extract ideal features that represents intrinsic characteristics of an image. Multiobject feature extraction system means a system that can extract features and locations of multiple objects in an image. In this paper we have discuss various methods to extract location and features of multiple objects and describe a system that can extract locations and features of multiple objects in an image by implementing an algorithm as hardware logic on a field-programmable gate array-based platform. There are many multiobject extraction methods which can be use for image segmentation based on motion, color intensity and texture. By calculating zeroth and first order moments of objects it is possible to obtain locations and sizes of multiple objects in an image.

In this paper, we describe the development of a system for evaluating an specific quality characteristic of recycled paper sheets using techniques of image analysis and pattern recognition. We call Bumpiness the phenomenon of interest, which is new in the literature on paper quality. This phenomenon is characterized by the appearance of macroscopic undulations on the paper sheet surface that may emerge shortly or some time after its production. We explore the detection and measurement of this defect by means of computer vision and statistical pattern recognition techniques that may allow early detection at the production site. Our goal is to give an scalar continuous measure of Bumpiness. We propose features computed from Gabor filter banks (GFB) and discrete wavelet transforms (DWT) for the characterization of paper sheet surface Bumpiness in recycled paper images. The starting point is to state the problem as a classification of the paper sheet images into two classes: low and high Bumpiness. In this setting we obtain, with both proposed texture modelling approaches (GFB and DWT), classification accuracies comparable to the agreement between human observers. The best performance is obtained using DWT features. Finally, we propose as the scalar index of Bumpines the fisher discriminant analysis (FDA) function defined on the space of the best features for the classification task. We perform an innovative validation process of this Bumpiness index, based on the ordering of random pairs of images, obtaining a very high agreement with the human observers.

紙幣や諸証券等のセキュリティ印刷物において，偽造品との差別化は重要な課題である。しかしながら，セキュリティ印刷物に新たな特徴的な要素を追加することは，コストの増加に繋がるため容易ではない。そこで，印刷物の基材となる「紙」がもつ情報を利用する分類手法に注目した。材料や製造条件等が異なることによって生まれる紙の特徴の違いを利用することが紙の分類において有効であると考えられる。現在までに，紙の分類方法として，紙の画像に現れる周期的な特徴の違いに注目した手法が提案され，その有効性が示されているが，より多様な紙の特徴の変化にも対応した分類方法として，紙の画像から算出可能な複数種類の特徴量を組み合せる方法が有効と考えられる。本報では，遺伝的アルゴリズム（Genetic Algorithm：GA）とサポートベクターマシン（Support Vector Machine：SVM）を組合せた分類器作成手法である特徴量選択型SVMを用いて，撮像された紙の画像から算出可能な複数種類の特徴量を基に学習を行い，材料や製造条件等が異なる紙の分類を行う分類器を自動的に構成する方法を提案する。分類器の構成に用いる画像特徴量は，階調値を用いた特徴量，2値画像から得られる特徴量，分割した画像から得られる特徴量及びGray Level Co―occurrence Matrix（GLCM）から得られる特徴量を用いた。提案手法の能力を確認するために，メーカーの異なる紙及び作製条件の異なる手すきシートを用いて分類器を構成する実験を行った。その結果，材料や制作方法の違い等，様々な条件が異なる紙に対応した高精度な分類器を自動的に構成することができ，提案手法の汎化性と有効性を確認することができた。

In this paper, a non-supervised technique for on-line paper characterisation is presented. The method uses self-organising maps (SOM) and texture analysis for clustering different kinds of paper according to their properties. A light-through technique is used to get pictures of paper. Then, effective texture features are extracted from greyscale images and the dimensionality of the feature data is reduced with SOM allowing visual analysis of measurements. The method makes it possible to implicitly extract important information about paper formation. The approach provides excellent results. A classification error below 1% was achieved for four quality classes when local binary pattern (LBP) texture features were used. The improvement to the previously used texture features in paper inspection is huge: the classification error was reduced by over 40 times. In addition to the excellent classification accuracy, the method also offers a self-intuitive user interface and a synthetic view of the inspected data.

For the recognition of paper currencies by image processing, the two steps data processing approach can yield high performance. The two steps include “recognition” and “verification” steps. In the current recognition machine, a simple statistical test is used as the verification step, where univariate Gaussian distribution is employed. Here we propose the use of the probability density formed by a multivariable Gaussian function, where the input data space is transferred to a lower dimensional subspace. Due to the structure of this model, we refer the total processing system as a hybrid neural network. Since the computation of the verification model only needs the inner product and square, the computational load is very small. In this paper, the method and numerical experimental results are shown by using the real data and the recognition machine

The nine award-winning papers in this special section were presented at the IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2010)that was held 13-18 June 2010 in San Francisco, CA.

Good registration (alignment to a reference) is essential for accurate face recognition. The effects of the number of landmarks on the mean localization error and the recognition performance are studied. Two landmarking methods are explored and compared for that purpose: (1) the most likely-landmark locator (MLLL), based on maximizing the likelihood ratio, and (2) Viola-Jones detection. Both use the locations of facial features (eyes, nose, mouth, etc) as landmarks. Further, a landmark-correction method (BILBO) based on projection into a subspace is introduced. The MLLL has been trained for locating 17 landmarks and the Viola-Jones method for 5. The mean localization errors and effects on the verification performance have been measured. It was found that on the eyes, the Viola-Jones detector is about 1% of the interocular distance more accurate than the MLLL-BILBO combination. On the nose and mouth, the MLLL-BILBO combination is about 0.5% of the inter-ocular distance more accurate than the Viola-Jones detector. Using more landmarks will result in lower equal-error rates, even when the landmarking is not so accurate. If the same landmarks are used, the most accurate landmarking method gives the best verification performance.

The three articles in this special section are selected papers from the IEEE CS Conference on Computer Vision and Pattern Recognition that was held in Anchorage, AL, in June 2008.

A deep learning neural network has been developed to efficiently and accurately clean the background noise in experimental lab-based X-ray diffraction images. A better spot segmentation is obtained and thus grain mapping in 3D is improved compared with the conventional method., Laboratory X-ray diffraction contrast tomography (LabDCT) is a novel imaging technique for non-destructive 3D characterization of grain structures. An accurate grain reconstruction critically relies on precise segmentation of diffraction spots in the LabDCT images. The conventional method utilizing various filters generally satisfies segmentation of sharp spots in the images, thereby serving as a standard routine, but it also very often leads to over or under segmentation of spots, especially those with low signal-to-noise ratios and/or small sizes. The standard routine also requires a fine tuning of the filtering parameters. To overcome these challenges, a deep learning neural network is presented to efficiently and accurately clean the background noise, thereby easing the spot segmentation. The deep learning network is first trained with input images, synthesized using a forward simulation model for LabDCT in combination with a generic approach to extract features of experimental backgrounds. Then, the network is applied to remove the background noise from experimental images measured under different geometrical conditions for different samples. Comparisons of both processed images and grain reconstructions show that the deep learning method outperforms the standard routine, demonstrating significantly better grain mapping.

Motivation Current spectral unmixing methods for multiplex fluorescence microscopy have a limited ability to cope with high spectral overlap as they only analyze spectral information over individual pixels. Here, we present adaptive Morphologically Constrained Spectral Unmixing (MCSU) algorithms that overcome this limitation by exploiting morphological differences between sub-cellular structures, and their local spatial context. Results The proposed method was effective at improving spectral unmixing performance by exploiting: (i) a set of dictionary-based models for object morphologies learned from the image data; and (ii) models of spatial context learned from the image data using a total variation algorithm. The method was evaluated on multi-spectral images of multiplex-labeled pancreatic ductal adenocarcinoma (PDAC) tissue samples. The former constraint ensures that neighbouring pixels correspond to morphologically similar structures, and the latter constraint ensures that neighbouring pixels have similar spectral signatures. The average Mean Squared Error (MSE) and Signal Reconstruction Error (SRE) ratio of the proposed method was 39.6% less and 8% more, respectively, compared to the best of all other algorithms that do not exploit these spatial constraints. Availability and Implementation Open source software (MATLAB). Supplementary information Supplementary data are available at Bioinformatics online.

Motivation Mutations (or Single Nucleotide Variants) in folded RiboNucleic Acid structures that cause local or global conformational change are riboSNitches. Predicting riboSNitches is challenging, as it requires making two, albeit related, structure predictions. The data most often used to experimentally validate riboSNitch predictions is Selective 2′ Hydroxyl Acylation by Primer Extension, or SHAPE. Experimentally establishing a riboSNitch requires the quantitative comparison of two SHAPE traces: wild-type (WT) and mutant. Historically, SHAPE data was collected on electropherograms and change in structure was evaluated by ‘gel gazing.’ SHAPE data is now routinely collected with next generation sequencing and/or capillary sequencers. We aim to establish a classifier capable of simulating human ‘gazing’ by identifying features of the SHAPE profile that human experts agree ‘looks’ like a riboSNitch. Results We find strong quantitative agreement between experts when RNA scientists ‘gaze’ at SHAPE data and identify riboSNitches. We identify dynamic time warping and seven other features predictive of the human consensus. The classSNitch classifier reported here accurately reproduces human consensus for 167 mutant/WT comparisons with an Area Under the Curve (AUC) above 0.8. When we analyze 2019 mutant traces for 17 different RNAs, we find that features of the WT SHAPE reactivity allow us to improve thermodynamic structure predictions of riboSNitches. This is significant, as accurate RNA structural analysis and prediction is likely to become an important aspect of precision medicine. Availability and Implementation The classSNitch R package is freely available at http://classsnitch.r-forge.r-project.org. Supplementary information Supplementary data are available at Bioinformatics online.

As of common routine in tumor resections, surgeons rely on local examinations of the removed tissues and on the swiftly made microscopy findings of the pathologist, which are based on intraoperatively taken tissue probes. This approach may imply an extended duration of the operation, increased effort for the medical staff, and longer occupancy of the operating room (OR). Mixed reality technologies, and particularly augmented reality, have already been applied in surgical scenarios with positive initial outcomes. Nonetheless, these methods have used manual or marker-based registration. In this work, we design an application for a marker-less registration of PET-CT information for a patient. The algorithm combines facial landmarks extracted from an RGB video stream, and the so-called Spatial-Mapping API provided by the HMD Microsoft HoloLens. The accuracy of the system is compared with a marker-based approach, and the opinions of field specialists have been collected during a demonstration. A survey based on the standard ISO-9241/110 has been designed for this purpose. The measurements show an average positioning error along the three axes of (x, y, z) = (3.3 ± 2.3, − 4.5 ± 2.9, − 9.3 ± 6.1) mm. Compared with the marker-based approach, this shows an increment of the positioning error of approx. 3 mm along two dimensions (x, y), which might be due to the absence of explicit markers. The application has been positively evaluated by the specialists; they have shown interest in continued further work and contributed to the development process with constructive criticism.

An automatic visual inspection system designed for dirt inspection in the pulp and paper industry is presented. A new hierarchical region oriented segmentation algorithm is introduced. The algorithm is tuned according to the singular characteristics of the pulp samples. A criterion based on the maximisation of the local contrast is defined in order to perform a defect region segmentation of the pulp and paper images. Some optimisations are introduced to avoid an excessive computational load. http://www.sciencedirect.com/science/article/B6V27-3XR260K-2T/1/aa6b716fc7193b3d447bff73516ce10d

A large variety of well-known scale-invariant texture recognition methods is tested with respect to their scale invariance. The scale invariance of these methods is estimated by comparing the results of two test setups. In the first test setup, the images of the training and evaluation set are acquired under same scale conditions and in the second test setup, the images in the evaluation set are gathered under different scale conditions than those of the training set. For the first test setup, scale invariance is not needed, whereas for the second test setup, scale invariance is obviously crucial. The difference between the results of these two test setups indicates the scale invariance of a method (the higher the scale invariance the lower the difference). The scale invariance of the methods is additionally estimated by analyzing the similarity of the feature vectors of images and their scaled versions. Additionally to the scale invariance, we also test eventual viewpoint and illumination invariance of the methods. As texture databases for our tests we use the KTH-TIPS database and the CUReT database. Results imply that many of the considered methods are not as scale-invariant as expected.

International audience; Heterogeneous image segmentation is one of the most important tasks in image processing. It consists in partitioning the image into a set of disjoint regions. In this paper, we propose a new unsupervised image segmentation method that we call Unsupervised Image Segmentation (UIS). Our proposal performs an efficient image partition efficiently into primitive regions. This process is ensured by a local adaptive Kmeans and a novel centroids initialization. Then, similar sets are agglomerated to form homogeneous regions. For that, a low-level feature merging is employed according to a hierarchical linkage approach. Finally, in case of over-segmentation, appearing outlier regions are removed using a post process stage. Therefore, the UIS method allows to determine automatically the image region number. Indeed, it extends the Kmeans clustering to obtain meaningful regions. Several experiments were conducted using two heterogeneous image datasets. A comparison with well-known segmentation methods was also performed using the Liu's factor measure.

The systems and concepts described in this paper document the evolution of the geometric invariance approach to object recognition over the last five years. Invariance overcomes one of the fundamental difficulties in recognising objects from images: that the appearance of an object depends on viewpoint. This problem is entirely avoided if the geometric description is unaffected by the imaging transformation. Such invariant descriptions can be measured from images without any prior knowledge of the position, orientation and calibration of the camera. These invariant measurements can be used to index a library of object models for recognition and provide a principled basis for the other stages of the recognition process such as feature grouping and hypothesis verification. Object models can be acquired directly from images, allowing efficient construction of model libraries without manual intervention. A significant part of the paper is a summary of recent results on the construction of invariants for 3D objects from a single perspective view. A proposed recognition architecture is described which enables the integration of multiple general object classes and provides a means for enforcing global scene consistency. Various criticisms of the invariant approach are articulated and addressed.

Bright field imaging of biological samples stained with antibodies and/or special stains provides a rapid protocol for visualizing various macromolecules. However, this method of sample staining and imaging is rarely employed for direct quantitative analysis due to variations in sample fixations, ambiguities introduced by color composition and the limited dynamic range of imaging instruments. We demonstrate that, through the decomposition of color signals, staining can be scored on a cell-by-cell basis. We have applied our method to fibroblasts grown from histologically normal breast tissue biopsies obtained from two distinct populations. Initially, nuclear regions are segmented through conversion of color images into gray scale, and detection of dark elliptic features. Subsequently, the strength of staining is quantified by a color decomposition model that is optimized by a graph cut algorithm. In rare cases where nuclear signal is significantly altered as a result of sample preparation, nuclear segmentation can be validated and corrected. Finally, segmented stained patterns are associated with each nuclear region following region-based tessellation. Compared to classical non-negative matrix factorization, proposed method: (i) improves color decomposition, (ii) has a better noise immunity, (iii) is more invariant to initial conditions and (iv) has a superior computing performance. contact: hchang@lbl.gov

Motivation: Caenorhabditis elegans, a roundworm found in soil, is a widely studied model organism with about 1000 cells in the adult. Producing high-resolution fluorescence images of C.elegans to reveal biological insights is becoming routine, motivating the development of advanced computational tools for analyzing the resulting image stacks. For example, worm bodies usually curve significantly in images. Thus one must ‘straighten’ the worms if they are to be compared under a canonical coordinate system. Results: We develop a worm straightening algorithm (WSA) that restacks cutting planes orthogonal to a ‘backbone’ that models the anterior–posterior axis of the worm. We formulate the backbone as a parametric cubic spline defined by a series of control points. We develop two methods for automatically determining the locations of the control points. Our experimental methods show that our approaches effectively straighten both 2D and 3D worm images. Contact: pengh@janelia.hhmi.org Supplementary information: The example data sets and programs are available upon request.

Motivation: A new technique, mammalian green fluorescence protein (GFP) reconstitution across synaptic partners (mGRASP), enables mapping mammalian synaptic connectivity with light microscopy. To characterize the locations and distribution of synapses in complex neuronal networks visualized by mGRASP, it is essential to detect mGRASP fluorescence signals with high accuracy. Results: We developed a fully automatic method for detecting mGRASP-labeled synapse puncta. By modeling each punctum as a Gaussian distribution, our method enables accurate detection even when puncta of varying size and shape partially overlap. The method consists of three stages: blob detection by global thresholding; blob separation by watershed; and punctum modeling by a variational Bayesian Gaussian mixture models. Extensive testing shows that the three-stage method improved detection accuracy markedly, and especially reduces under-segmentation. The method provides a goodness-of-fit score for each detected punctum, allowing efficient error detection. We applied this advantage to also develop an efficient interactive method for correcting errors. Availability: The software is available on http://jinny.kist.re.kr Contact: tingzhao@gmail.com; kimj@kist.re.kr

PURPOSE: In functional MRI (fMRI), faster sampling of data can provide richer temporal information and increase temporal degrees of freedom. However, acceleration is generally performed on a volume-by-volume basis, without consideration of the intrinsic spatio-temporal data structure. We present a novel method for accelerating fMRI data acquisition, k-t FASTER (FMRI Accelerated in Space-time via Truncation of Effective Rank), which exploits the low-rank structure of fMRI data. THEORY AND METHODS: Using matrix completion, 4.27× retrospectively and prospectively under-sampled data were reconstructed (coil-independently) using an iterative nonlinear algorithm, and compared with several different reconstruction strategies. Matrix reconstruction error was evaluated; a dual regression analysis was performed to determine fidelity of recovered fMRI resting state networks (RSNs). RESULTS: The retrospective sampling data showed that k-t FASTER produced the lowest error, approximately 3-4%, and the highest quality RSNs. These results were validated in prospectively under-sampled experiments, with k-t FASTER producing better identification of RSNs than fully sampled acquisitions of the same duration. CONCLUSION: With k-t FASTER, incoherently under-sampled fMRI data can be robustly recovered using only rank constraints. This technique can be used to improve the speed of fMRI sampling, particularly for multivariate analyses such as temporal independent component analysis.

Evaluation of previous systems for automated determination of subcellular location from microscope images has been done using datasets in which each location class consisted of multiple images of the same representative protein. Here, we frame a more challenging and useful problem where previously unseen proteins are to be classified.Using CD-tagging, we generated two new image datasets for evaluation of this problem, which contain several different proteins for each location class. Evaluation of previous methods on these new datasets showed that it is much harder to train a classifier that generalizes across different proteins than one that simply recognizes a protein it was trained on. We therefore developed and evaluated additional approaches, incorporating novel modifications of local features techniques. These extended the notion of local features to exploit both the protein image and any reference markers that were imaged in parallel. With these, we obtained a large accuracy improvement in our new datasets over existing methods. Additionally, these features help achieve classification improvements for other previously studied datasets.The datasets are available for download at http://murphylab.web.cmu.edu/data/. The software was written in Python and C++ and is available under an open-source license at http://murphylab.web.cmu.edu/software/. The code is split into a library, which can be easily reused for other data and a small driver script for reproducing all results presented here. A step-by-step tutorial on applying the methods to new datasets is also available at that address.murphy@cmu.eduSupplementary data are available at Bioinformatics online.

Oblique images are aerial photographs taken at oblique angles to the earth's surface. Projections of vector and other geospatial data in these images depend on camera parameters, positions of the geospatial entities, surface terrain, occlusions, and visibility. This paper presents a robust and scalable algorithm to detect inconsistencies in vector data using oblique images. The algorithm uses image descriptors to encode the local appearance of a geospatial entity in images. These image descriptors combine color, pixel-intensity gradients, texture, and steerable filter responses. A Support Vector Machine classifier is trained to detect image descriptors that are not consistent with underlying vector data, digital elevation maps, building models, and camera parameters. In this paper, we train the classifier on visible road segments and non-road data. Thereafter, the trained classifier detects inconsistencies in vectors, which include both occluded and misaligned road segments. The consistent road segments validate our vector, DEM, and 3-D model data for those areas while inconsistent segments point out errors. We further show that a search for descriptors that are consistent with visible road segments in the neighborhood of a misaligned road yields the desired road alignment that is consistent with pixels in the image., In Proceedings of 16th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems (ACM GIS'08)

Surveillance cameras are everywhere, keeping an eye on pedestrians as they navigate through a scene. With this context, our paper addresses the problem of pedestrian attribute recognition (PAR). This problem entails recognizing attributes such as age-group, clothing style, accessories, footwear style etc. This is a multi-label problem and challenging even for human observers. The problem has rightly attracted attention recently from the computer vision community. In this paper, we adopt trainable Gabor wavelets (TGW) layers and use it with a convolution neural network (CNN). Whereas other researchers are using fixed Gabor filters with the CNN, the proposed layers are learnable and adapt to the dataset for a better recognition. We propose a multi-branch neural network where mixed-layers, a combination of the TGW and convolutional layer, make up the building block of our 3-branch deep neural network. We test our method on publicly available challenging datasets and compare our results with state of the art.

This paper proposes the use of a discrete cosine transform (DCT) instead of the eigenfaces method (Karhunen-Loeve Transform) for biometric identification based on frontal face images. Experimental results show better recognition accuracies and reduced computational burden. This paper includes results with different classifiers and a combination of them., Comment: 9 pages, published in IEEE 37th Annual 2003 International Carnahan Conference on Security Technology, 2003. Proceedings. 14-16 Oct. 2003 Taipei (Taiwan)

Aim: This paper describes the digital implementation of a mathematical transform namely 2D Fast Discrete Curvelet Transform (FDCT) via UnequiSpaced Fast Fourier Transform (USFFT) in combination with the novel segmentation method for effective detection of breast cancer. Methods: USFFT performs exact reconstructions with high image clarity. Radon, ridgelet and Cartesian filters are included in this method. Mean Square Error (MSE) and Peak Signal to Noise Ratio (PSNR) were calculated for the image and the resulting value showed that the proposed method performs well on mammogram image in reducing noise with good extraction of edges. This work includes a novel segmentation method, which combines Modified Local Range Modification (MLRM) and Laplacian of Gaussian (LoG) edge detection method to segment the textured features in the mammogram image. Results: The result was analyzed using a Receiver Operating Characteristics (ROC) plot and the detection accuracy found was 99% which is good compared to existing methods.

PurposeThe gradual increase in geriatric issues and global imbalance of the ratio between patients and healthcare professionals have created a demand for intelligent systems with the least error-prone diagnosis results to be used by less medically trained persons and save clinical time. This paper aims at investigating the development of image-based colourimetric analysis. The purpose of recognising such tests is to support wider users to begin a colourimetric test to be used at homecare settings, telepathology and so on.Design/methodology/approachThe concept of an automatic colourimetric assay detection is delivered by utilising two cases. Training deep learning (DL) models on thousands of images of these tests using transfer learning, this paper (1) classifies the type of the assay and (2) classifies the colourimetric results.FindingsThis paper demonstrated that the assay type can be recognised using DL techniques with 100% accuracy within a fraction of a second. Some of the advantages of the pre-trained model over the calibration-based approach are robustness, readiness and suitability to deploy for similar applications within a shorter period of time.Originality/valueTo the best of the authors’ knowledge, this is the first attempt to provide colourimetric assay type classification (CATC) using DL. Humans are capable to learn thousands of visual classifications in their life. Object recognition may be a trivial task for humans, due to photometric and geometric variabilities along with the high degree of intra-class variabilities, it can be a challenging task for machines. However, transforming visual knowledge into machines, as proposed, can support non-experts to better manage their health and reduce some of the burdens on experts.

Deep Convolutional Neural Networks (CNNs) can be overly dependent on training data, causing a generalization problem in which trained models may not predict real-world datasets. To address this problem, various regularization methods such as image manipulation and feature map regularization have been proposed for their strong generalization ability. In this paper, we propose a regularization method that applies both image manipulation and feature map regularization based on patches. The method proposed in this paper has a regularization effect in two stages, which makes it possible to better generalize the model. Consequently, it improves the performance of the model. Moreover, our method adds features extracted from other images in the hidden state stage, which not only makes the model robust to noise but also captures the distribution of each label. Through experiments, we show that our method performs competently on models that generate a large number of parameter and multiple feature maps for the CIFAR and Tiny-ImageNet datasets.

A novel classifier for face recognition using an improved probabilistic collaborative representation named IPCR is proposed in this paper. The purpose of this paper is to improve the accuracy of face recognition. The testing sample is assumed to be linearly combined by a part of training samples in feature space. There is two-phase framework in IPCR. In the first phase, an adjusted parameter of the nearest neighbors of the samples is chosen for classification. In the second phase, a linear combination of the features and the sparse coefficients are used for new patterns. In the process of two-phase framework, the weight matrix is obtained according to the distance between all the training samples and each testing sample, and then it is applied to weight probabilistic collaborative representation coefficients. The kernel trick is implemented for the high-dimensional nonlinear information instead of linear information of data to improve the class separability. The second classifier named KPCR uses a kernel probabilistic collaborative representation for face recognition. Several renowned face databases, e.g., AR, GT, PIE, FERET, and LFW-crop are used for evaluating the performances of the proposed classifiers. The experimental results demonstrate that the proposed classifiers outperform the collaborative representation-based classification (CRC), the probabilistic collaborative representation-based classifier (ProCRC), and the other state-of-the-art classifiers in recognition accuracy.

This paper presents a low power near real-time pattern recognition technique based on Mathematical Morphology-MM implemented on FPGA (Field Programmable Gate Array). The key to the success of this approach concerns the advantages of machine learning paradigm applied to the translation invariant template-matching operators from MM. The paper shows that compositions of simple elementary operators from Mathematical Morphology based on ELUTs (Elementary Look-Up Tables) are very suitable to embed in FPGA hardware. The paper also shows the development techniques regarding all mathematical modeling for computer simulation and system generating models applied for hardware implementation using FPGA chip. In general, image processing on FPGAs requires low-level description of desired operations through Hardware Description Language-HDL, which uses high complexity to describe image operations at pixel level. However, this work presents a reconfiguring pattern recognition device implemented directly in FPGA from mathematical modeling simulation under Matlab/Simulink/System Generator environment. This strategy has reduced the hardware development complexity. The device will be useful mainly when applied on remote sensing tasks for aerospace missions using passive or active sensors. El presente trabajo presenta una técnica de reconocimiento de patrones en tiempo real basada en Morfología Matemática-MM implementada en FPGA (Field Programmable Gate Array). La estrategia para la efectividad de este enfoque tiene que ver con las ventajas del paradigma de aprendizaje automática aplicada al modelo de correspondencia con la invariancia traslacional de operadores elementales da MM. El artículo muestra que las composiciones de operadores elementales simples de morfología matemática basadas en ELUT (tablas de consulta elementales) son adecuadas para integrarse en dispositivos FPGA. Este artículo también muestra técnicas de desarrollo de sistemas de reconocimiento de patrones, desde el modelado matemático de operadores morfológicos hasta la implementación del dispositivo electrónico utilizando el software System Generator. En general, las operaciones para el procesamiento de imágenes en FPGAs se implementan a un bajo nivel de abstracción de los lenguajes de descripción del hardware-HDL. Esto crea una gran complejidad en la implementación de operaciones en imágenes a nivel de píxeles. Sin embargo, este trabajo presenta un dispositivo reconfigurable de reconocimiento de patrones implementado directamente en FPGA a partir de simulación de modelado matemático en el software Matlab/Simulink/System Generator. Esta estrategia ha reducido la complejidad del desarrollo de hardware. El dispositivo será útil principalmente cuando se aplique en tareas de teledetección para misiones aeroespaciales utilizando sensores pasivos o activos. Esse trabalho apresenta uma técnica de reconhecimento de padrões baseada em Morfologia Matemática-MM, implementada em FPGA (Field Programmable Gate Array). A estratégia para o êxito dessa abordagem consiste na utilização das vantagens do paradigma de aprendizagem de máquina aplicado em operadores morfológicos de casamento de padrões invariantes à translação. Esse artigo mostra que a composição de simples operadores elementares da MM baseados em ELUTS (Elementary Look-Up Tables) são adequados para aplicações embarcadas em FPGA. Esse artigo também mostra as técnicas de desenvolvimento do sistema de reconhecimento de padrões, desde a modelagem matemática dos operadores morfológicos até a implementação do dispositivo eletrônico usando o software System Generator. Em geral, as operações para o processamento de imagens em FPGAs são implementadas em baixo nível de abstração das linguagens de descrição de hardware-HDL. Isto gera alta complexidade na implementação de operações em imagens ao nível de pixel. No entanto, esse trabalho apresenta um dispositivo reconfigurável aplicado ao reconhecimento de padrões implementado em FPGA, a partir da simulação da modelagem matemática usando o ambiente de software Matlab/Simulink/System Generator. Essa estratégia reduz a complexidade do desenvolvimento em hardware. O dispositivo apresentado deverá ser útil principalmente quando aplicado em tarefas de sensoriamento remoto para missões aeroespaciais através de sensores passivos ou ativos.

Table detection is a preliminary step in extracting reliable information from tables in scanned document images. We present CasTabDetectoRS, a novel end-to-end trainable table detection framework that operates on Cascade Mask R-CNN, including Recursive Feature Pyramid network and Switchable Atrous Convolution in the existing backbone architecture. By utilizing a comparativelyightweight backbone of ResNet-50, this paper demonstrates that superior results are attainable without relying on pre- and post-processing methods, heavier backbone networks (ResNet-101, ResNeXt-152), and memory-intensive deformable convolutions. We evaluate the proposed approach on five different publicly available table detection datasets. Our CasTabDetectoRS outperforms the previous state-of-the-art results on four datasets (ICDAR-19, TableBank, UNLV, and Marmot) and accomplishes comparable results on ICDAR-17 POD. Upon comparing with previous state-of-the-art results, we obtain a significant relative error reduction of 56.36%, 20%, 4.5%, and 3.5% on the datasets of ICDAR-19, TableBank, UNLV, and Marmot, respectively. Furthermore, this paper sets a new benchmark by performing exhaustive cross-datasets evaluations to exhibit the generalization capabilities of the proposed method.

The emergence of the novel coronavirus pneumonia (Covid-19) pandemic at the end of 2019 led to chaos worldwide. The world breathed a sigh of relief when some countries announced that they had obtained the appropriate vaccine and gradually began to distribute it. Nevertheless, the emergence of another wave of this disease has returned us to the starting point. At present, early detection of infected cases has been the paramount concern of both specialists and health researchers. This paper aims to detect infected patients through chest x-ray images. The large dataset available online for Covid-19 (COVIDx) was used in this research. The dataset consists of 2,128 x-ray images of Covid-19 cases, 8,066 normal cases, and 5,575 cases of pneumonia. A hybrid algorithm was applied to improve image quality before conducting the neural network training process. This algorithm consisted of combining two different noise reduction filters in the images, followed by a contrast enhancement algorithm. In this paper, for Covid-19 detection, a novel convolution neural network (CNN) architecture, KL-MOB (Covid-19 detection network based on MobileNet structure), was proposed. KL-MOB performance was boosted by adding the Kullback–Leibler (KL) divergence loss function at the end when trained from scratch. The Kullback–Leibler (KL) divergence loss function was adopted as content-based image retrieval and fine-grained classification to improve the quality of image representation. This paper yielded impressive results, overall benchmark accuracy, sensitivity, specificity, and precision of 98.7%, 98.32%, 98.82%, and 98.37%, respectively. The promising results in this research may enable other researchers to develop modern and innovative methods to aid specialists. The tremendous potential of the method proposed in this research can also be utilized to detect Covid-19 quickly and safely in patients throughout the world.

When an intense brightness for a small amount of time is seen in the sun, then we can say that a solar flare emerged. As solar flares are made up of high energy photons and particles, thus causing the production of high electric fields and currents and therefore results in the disruption in space-borne or ground-based technological system. It also becomes a challenging task to extract its important features for prediction. Convolutional Neural Networks have gain a significant amount of popularity in the classification and localization tasks. This paper has given stress on the classification of the solar flares emerged on different years by stacking different convolutional layers followed by max pooling layers. From the reference of Alexnet, the pooling layer employed in this paper is the overlapping pooling. Also two different activation functions that are ELU and CReLU have been used to investigate how many number of convolutional layers with a particular activation function provides with the best results on this dataset as the size of the dataset in this domain is always small. The proposed investigation can be further used in a novel solar prediction systems.

Keeping an eye on pedestrians as they navigate through a scene, surveillance cameras are everywhere. With this context, our paper addresses the problem of pedestrian attribute recognition (PAR). This problem entails recognizing attributes such as age-group, clothing style, accessories, footwear style etc. This multi-label problem is extremely challenging even for human observers and has rightly garnered attention from the computer vision community. Towards a solution to this problem, in this paper, we adopt trainable Gabor wavelets (TGW) layers and cascade them with a convolution neural network (CNN). Whereas other researchers are using fixed Gabor filters with the CNN, the proposed layers are learnable and adapt to the dataset for a better recognition. We propose a two-branch neural network where mixed layers, a combination of the TGW and convolutional layers, make up the building block of our deep neural network. We test our method on twoo challenging publicly available datasets and compare our results with state of the art.

The latest research has shown that adaptive directional wavelet transform can constitute edges and textures in images efficiently due to the adaptive directional selectivity. This paper is primarily focused on the application of adaptive directional wavelet transform in conjunction with linear discriminant analysis (LDA) for capturing the discriminant directional multiresolution facial features. The intention of this paper is to explore the efficacy of adaptive directional wavelet transform in facial feature extraction and to offer a stepping stone for further research in this direction. The proposed approach is compared with existing subspace and local descriptor feature extraction methods. A performance comparison is also demonstrated with existing non-adaptive multiresolution analysis methods such as discrete wavelet transform (DWT), Gabor wavelet transform (GWT), curvelets, ridgelets, contourlets, and local Gabor binary pattern. Evaluation of the proposed approach on famous databases such as ORL, Essex Grimace, Yale, and Sterling face convinces the effectiveness of the adaptive directional wavelet transform based subspace features. Keywords: Face recognition, Adaptive directional wavelet transform, Linear discriminant analysis, Multiresolution analysis

In this paper, we present a new hashing method to learn compact binary codes for highly efficient image retrieval on large-scale datasets. While the complex image appearance variations still pose a great challenge to reliable retrieval, in light of the recent progress of Convolutional Neural Networks (CNNs) in learning robust image representation on various vision tasks, this paper proposes a novel Deep Supervised Hashing method to learn compact similarity-preserving binary code for the huge body of image data. Specifically, we devise a CNN architecture that takes pairs/triplets of images as training inputs and encourages the output of each image to approximate discrete values (e.g. $$+\,1$$/$$-\,1$$). To this end, the loss functions are elaborately designed to maximize the discriminability of the output space by encoding the supervised information from the input image pairs/triplets, and simultaneously imposing regularization on the real-valued outputs to approximate the desired discrete values. For image retrieval, new-coming query images can be easily encoded by forward propagating through the network and then quantizing the network outputs to binary codes representation. Extensive experiments on three large scale datasets CIFAR-10, NUS-WIDE, and SVHN show the promising performance of our method compared with the state-of-the-arts.

Domain adaptation (DA), a particular case of transfer learning, is an effective technology for learning a discriminative model in scenarios where the data from the training (source) and the testing (target) domains share common class labels but follow different distributions. The differences between domains, called domain shifts, are caused by variations in the acquisition devices and environmental conditions, such as changing illuminations, pose, and collecting-device noises, that are related to a specific domain, denoted as domain-specific noises in this paper. The research on stacked denoising autoencoder (SDA) has demonstrated that noise-robust features could be learned through training a model to reduce the man-made (simulated) noises. However, little research has been conducted to learn the domain-invariant features through training SDA to reduce the domain-specific noises from the real word. In this paper, we propose a novel variant of SDA for DA, called the stacked local constraint auto-encoder (SLC-AE), which aims to learn domain-invariant features through iteratively optimizing the SDA and the low-dimensional manifold. The core idea behind the SLC-AE is that both the source and target samples are corrupted due to the domain-specific noises, and each corrupted sample could be de-noised by calculating the weighted sum of its neighbor samples defined on the intrinsic manifold. Because the neighbor samples on the intrinsic manifold are semantically similar, their weighted sum preserves the generic information and reduces the domain-specific noises. To properly evaluate the performance of the SLC-AE, we conducted extensive experiments using seven benchmark data sets, i.e., MNIST, USPS, COIL20, SYN SIGNS, GTSRB, MSRC and VOC 2007. Compared to twelve different state-of-the-art methods, the experimental results demonstrated that the proposed SLC-AE model made significant improvement over the performance of SDA and achieved the best average performance on the seven data sets.

With the development of video surveillance in public safety field, there is an increasing research on person re-identification (re-id). In this paper, we address the image-to-video person re-id, in which the probe is an image and the gallery is consists of videos captured by nonoverlapping cameras. Compared with image, video sequence contains more temporal information that can be explored to improve the performance of re-identification system. However, it is challenging to model temporal information in the matching process of image-to-video person re-id. In this paper, we proposed a novel temporally memorized similarity learning neural network for this problem. In specific, the proposed network mainly consisted of two parts, including feature representation sub-network and similarity sub-network. In the first part, we adopted a convolutional neural network (CNN) to extract features from the input image. Given a video sequence of a person, features were first extracted from each its frame by using CNN and further forward to a long shot term memory (LSTM) network to encode the temporal information of video sequence. The outputs of LSTM were concatenated together as the feature vector of video sequences. Finally, the feature vectors of probe image and the video sequence were further forward to the similarity sub-network for distance metric learning. In the proposed framework, the feature representation and the similarity metric learning can be learned and optimized simultaneously. We evaluated the proposed framework on three public person re-id data sets, and the experimental results showed that the proposed approach is effective for the image-to-video person re-id.

This paper proposes methods for super resolving single noisy low resolution images. Even if single image super resolution has been a topic of research for last few decades, super resolution of noisy low resolution images is still a challenging problem. Most of the state of the art super resolution algorithms will fail to perform if significant amount of noise is present in the observed image. In this paper, we propose a denoised patch dictionary based single image super resolution algorithm. To enhance the robustness to noise performance, this method is further modified by using a wavelet based fusion algorithm which combines the result of proposed method with direct super resolved image, and super resolved image after denoising to preserve the finer details of the super resolved image. The proposed methods are applied on the commonly used test images. The results validate that the proposed methods show improvement over the existing techniques.

In this paper, the problem of human action recognition using 3D reconstruction data is deeply investigated. 3D reconstruction techniques are employed for addressing two of the most challenging issues related to human action recognition in the general case, namely view variance (i.e., when the same action is observed from different viewpoints) and the presence of (self-) occlusions (i.e., when for a given point of view a body part of an individual conceals another body part of the same or another subject). The main contributions of this paper are summarized as follows. The first one is a detailed examination of the use of 3D reconstruction data for performing human action recognition. The latter includes the introduction of appropriate local/global flow/shape descriptors, extensive experiments in challenging publicly available datasets and exhaustive comparisons with state-of-art approaches. The second one is a new local-level 3D flow descriptor, which incorporates spatial and surface information in the flow representation and efficiently handles the problem of defining 3D orientation at every local neighborhood. The third one is a new global-level 3D flow descriptor that efficiently encodes the global motion characteristics in a compact way. The fourth one is a novel global temporal-shape descriptor that extends the notion of 3D shape descriptions for action recognition, by incorporating the temporal dimension. The proposed descriptor efficiently addresses the inherent problems of temporal alignment and compact representation, while also being robust in the presence of noise (compared with similar tracking-based methods of the literature). Overall, this paper significantly improves the state-of-art performance and introduces new research directions in the field of 3D action recognition, following the recent development and wide-spread use of portable, affordable, high-quality and accurate motion capturing devices (e.g., Microsoft Kinect).