Transformer has shown outstanding performance on image denoising, but the existing Transformer methods for image denoising are with large model sizes and high computational complexity, which is unfriendly to resource-constrained devices. In this paper, we propose a Lightweight Image Denoising Transformer method (LIDFormer) based on Triple Multi-Dconv Head Transposed Attention (TMDTA) to boost computational efficiency. LIDFormer first implements Discrete Wavelet Transform (DWT), which transforms the input image into a low-frequency space, greatly reducing the computational complexity of image denoising. However, the low-frequency image lacks fine-feature information, which degrades the denoising performance. To handle this problem, we introduce the Complementary Periodic Feature Reusing (CPFR) scheme for aggregating the shallow-layer features and the deep-layer features. Furthermore, TMDTA is proposed to integrate global context along three dimensions, thereby enhancing the ability of global feature representation. Note that our method can be applied as a pipeline for both convolutional neural networks and Transformers. Extensive experiments on several benchmarks demonstrate that the proposed LIDFormer achieves a better trade-off between high performance and low computational complexity on real-world image denoising tasks. 

pdf/efficient-lightweight-image-denoising-with-triple-attention-ylp0a9uhtu.pdf

Efficient Lightweight Image Denoising with Triple Attention Transformer

Esta tesis presenta un nuevo enfoque para estimar cantidades cinematicas, a saber, el angulo de la velocidad relativa y la aceleracion de una imagen de desenfoque de movimiento unico real usando la Transformacion discreta de coseno (DCT). El desenfoque de movimiento es un fenomeno comun presente en las imagenes. Se produce por el movimiento relativo entre la camara y los objetos en la escena durante la exposicion del sensor de la camara a la luz. Suele ocurrir con los sistemas de grabacion de imagenes montados en vehiculos, camaras de mano, drones, satelites y robots moviles. La presente tecnica basada en software se enfoca en casos donde la camara se mueve a una velocidad lineal constante mientras el fondo permanece sin cambios. Para los experimentos de estimacion de velocidad se usaron imagenes sinteticas y reales. El error absoluto medio (MAE) del metodo DCT Radon para la estimacion de direccion fue de 4.66 grados. Ademas, el error relativo medio para la estimacion de la velocidad del DCT Pseudo Cepstrum fue del 5.15%. Las propuestas alternativas de analisis de frecuencia DCT fueron mas precisas que todos los competidores evaluados para la medicion de velocidad. Adicionalmente, se propuso un enfoque alternativo para estimar la aceleracion relativa a partir de una imagen de desenfoque de movimiento acelerado uniformemente real usando mapeo homomorfico para extraer la funcion de dispersion de puntos caracteristica de una imagen degradada para luego entrenar un modelo de regresion de aprendizaje automatico. Se tomaron un total de 125 imagenes de desenfoque de movimiento uniformemente acelerado en un entorno controlado con luz y distancia a 5 aceleraciones diferentes en un rango entre0.64m/s2 y2.4m/s2. Se evaluaron Conjuntos de arboles, procesos gaussianos (GPR), Regresion Lineal, Maquinas de Vectores de Soporte (SVM) y 19 variantes de regresion para predecir la aceleracion. El mejor resultado RMSE fue de 0.2553m/s2 usando regresion GPR con una velocidad de prediccion de 530 observaciones por segundo. Finalmente, se concluye que los metodos propuestos son alternativas validas para la estimacion de la velocidad y la aceleracion de una sola imagen con desenfoque de movimiento lineal invariante.

A Contribution to the Estimation of Kinematic Quantities from Linear Motion Blurred Images

Machine learning and deep learning classiﬁcation models are data-driven, and the model and the data jointly determine their classiﬁcation performance. It is biased to evaluate the model’s performance only based on the classiﬁer accuracy while ignoring the data separability. Sometimes, the model exhibits excellent accuracy, which might be attributed to its test-ing on highly separable data. Most of the current studies on data separability measures are deﬁned based on the distance between sample points, but this has been demonstrated to fail in several circumstances. In this paper, we propose a new separability measure—the rate of separability (RS), which is based on the data coding rate. We validate its eﬀectiveness as a supplement to the separability measure by comparing it to four other distance-based measures on synthetic datasets. Then, we demonstrate the positive correlation between the proposed measure and recognition accuracy in a multi-task scenario constructed from a real dataset. Finally, we discuss the methods for evaluating the classiﬁcation performance of machine learning and deep learning models considering data separability.

A classification performance evaluation measure considering data separability

Rail crack is a serious threat to the safety of train operations. Accurately classifying various types of rail crack is a crucial aspect in preventing potential accidents. To address this challenge, this paper proposes an improved rail crack classification method based on wavelet multi-resolution analysis (WMRA) and signal separability. By using electromagnetic acoustic emission (EMAE) technique, crack-related acoustic emission (AE) signals are generated, which are then decomposed into sub-signals using WMRA. Subsequently, the best feature set (BFS) is obtained from each sub-signal based on the constructed feature separability index (ST). The signal separability index (SSI) is constructed to evaluate the information quantity of each sub-signal for subsequent signal analysis. Finally, by leveraging the BFS and the most informative sub-signal, crack classification is performed. Comparative analysis with conventional approaches reveals a remarkable enhancement in the accuracy of rail crack classification, specifically within a reduced feature dimension. Noteworthy, this method demonstrates the frequency band concentration effect of EMAE, thereby providing valuable insights for further research in the field of rail crack classification.

An Improved Rail Crack Classification Method Based on Wavelet Multi-Resolution Analysis and Signal Separability

. © The author; licensee Universidad de la Costa CUC. INGE CUC vol. 16 no. 1, pp. 80-94. Enero Junio, 2020 Barranquilla. ISSN 0122-6517 Impreso, ISSN 2382-4700 Online . Resumen Introducción− En este artículo se presenta el diseño y la construcción de un riel electromecánico que permite obtener la velocidad y aceleración instantáneas de una plataforma que soporta una cámara científica al momento de la captura de la imagen, para el estudio de las imágenes difuminadas por movimiento uniforme (Motion blur). Se presentan los requerimientos de diseño mecánico y eléctrico basados en las condiciones de presentación del fenómeno. Seguido, el sistema fue calibrado con respecto a un instrumento patrón para el cálculo de incertidumbres y error. Objetivo− En este artículo se presenta el diseño, construcción y calibración de un sistema electromecánico para el estudio de las imágenes con difuminación uniforme. Metodología− El desarrollo del sistema se dividió en los siguientes pasos: El diseño y la construcción de un riel electromecánico en aluminio con una plataforma móvil de velocidad constante; la calibración de la velocidad de la plataforma móvil utilizando la Guía para la estimar la incertidumbre de la medición (GUM); y el diseño y la construcción de un riel electromecánico en aluminio con una plataforma móvil de aceleración constante y calibración de la aceleración de la plataforma móvil utilizando la Guía para la estimar la incertidumbre de la medición (GUM). Resultados− Se obtuvieron incertidumbres máximas de para la velocidad y de para la aceleración del sistema. Conclusiones− El sistema desarrollado corresponde a un sistema electromecánico que permite desplazar una plataforma con cámara a lo largo de un par de barras paralelas, de baja incertidumbre con la posibilidad de medir la velocidad y aceleración instantáneas para el estudio de imágenes reales con difuminadas por movimiento uniforme y la enseñanza de física básica. Palabras clave− Cinemática; máquinas electrónicas; metrología; prototipo; riel Abstract Introduction− This paper introduces the design and the construction of an electromechanical slider that allows obtaining the instantaneous speed and acceleration of a platform that holds a scientific camera to take photos for the study motion-blurred images. The mechanical and electrical design requirements are given based on the conditions of presentation of the phenomenon. The system was calibrated concerning a standard instrument for the estimation of the uncertainty and error. Objective− In this paper, the design, construction and calibration of an electromechanical system for the study of images with uniform motion blur is presented. Methodology− The development of the system is split into the following steps: The design and construction of an aluminum electromechanical slider built with a mobile platform that moves at constant speed; the calibration of the speed of the mobile platform using the Guide to Estimate Uncertainty in the Measurement (GUM); and the design and construction of the electromechanical aluminum slider mounted with a mobile platform at constant acceleration and the calibration of the mobile platform for the acceleration applying the Guide for Estimation of Measurement Uncertainty (GUM). Results− Maximum uncertainties of were obtained for speed and for acceleration of the system. Conclusions− The developed system corresponds to an electromechanical system that allows to move a cart along a pair of parallel bars, of low uncertainty with the possibility of measuring instantaneous speed and acceleration for the study of motion blurred images and teaching of motion physics. Keywords− Slider; electronic machines; metrology; kinematics; prototype •

/pdf/diseno-y-construccion-de-un-riel-electromecanico-para-el-ish1nn40im.pdf

Diseño y construcción de un riel electromecánico para el estudio de la cinemática de imágenes con difuminación lineal uniforme

Feature extraction is a key step in the classification and recognition problem. Features from different methods vary a lot with different separability in their feature space. We propose a novel method based on the distance matrix to evaluate feature separability by describing the in-class aggregation and the between-class scatter of every class. Finally the separability of each feature class is measured individually. Experiments on the synthetic data and ORL face dataset prove its effectiveness and advantage with regard to the conventional methods.

Feature separability based on the distance matrix

This paper we propose an novel parametric approach for single image kernel estimation with both motion blur and Gaussian blur coupled. In the view of that daily pictures captured by handheld device usually contain motion blur and defocus simultaneously. During one shot, the moving trail of the object can be always regarded as straight and consecutive, and the defocus phenomenon is related to Gaussian blur. Therefore, a parameter model containing three parameters can describe the blur. First, we estimate a rough blur kernel using L 1 prior method, then we fit the kernel by computing the three parameters. Finally, the sharp image with clear details is restored by the kernel estimated. Experimental results show that the proposed method outperforms others when the blur kernel is fairly parameterized, which helps the current blind deconvolution methods achieve better results.

Parametric model for image blur kernel estimation

Image denoising is a fundamental but critical task. Previous works based on deep networks have made great progress, but suffer from the problem of computational overload. This paper addresses the demands by (1) a lightweight denoising network and (2) a novel knowledge distillation algorithm. The experimental results show the usefulness of the RS-KD on the proposed lightweight network and consistent gains that can be obtained on both synthetic and real-world datasets. Especially, benefiting from the retargeting supervision, our proposed distillation framework allows for arbitrary high-performance teacher networks. 

Toward Efficient Image Denoising: A Lightweight Network with Retargeting Supervision Driven Knowledge Distillation

Surveillance video anomaly detection is a challenging problem because of the diversity of abnormal events. The current prediction-based methods outperform reconstruction-based methods. But the former has the following issues: 1) Using optical flow to represent motion will affect real-time detection. 2) Distinguishing abnormal events only by local relationships will lead to ambiguity. 3) Semantic information and spatiotemporal constraint are not fully utilized. To address these problems, we propose FECP-Net: a network with feature enhancement and consistency frame prediction for surveillance video anomaly detection. We use the RGB difference between consecutive frames rather than optical flow to realize real-time detection. Meanwhile, we design a feature enhancement module to enrich semantics and global context information in features. In addition, we add spatiotemporal consistency constraint and consistency loss to strengthen consistency predictions. Extensive experiments on standard benchmarks demonstrate the effectiveness of our method.

Surveillance Video Anomaly Detection with Feature Enhancement and Consistency Frame Prediction

Handle everyday research tasks with reliable, citation-backed results

Your personal Research Agent to handle research tasks with citation-backed results

Popular Tasks used by Researchers

How can I help with your research?

Meet SciSpace

Get more enhanced response by uploading the PDFs you want me to reference.

No relevant PDFs in your library

SciSpace is the AI research assistant for academics. Run systematic literature reviews on 280M+ papers, and write papers with cited sources. Trusted by 1M+ students, PhDs & researchers.

SciSpace | AI for Research

Analyze PDFs

Code & Manuscripts

Funding & Grants

Literature & Patents

Medical & Clinical Data

Systematic Review

Visualize & Present

Web & Data

Build a Google Scholar-like website for your research.

Build a website

Create charts and images for your research

Create a Chart

Write a paper for submission to a journal

Draft a manuscript

Patent Search

Design eye-catching scientific posters in minutes.

Scientific Poster Generation

Systematic Literature Review

One task is running at the moment. Your messages will be shown right after.

Drag and drop or click here to browse

Loved by <highlight>1 million+</highlight> researchers

Extract a list of specific topics and their sources from unstructured text

Topics

Compare and analyze relevant papers that matches with your search

Papers

Get insights from PDFs and bookmarked papers from your library

My library

Recent searches

Try searching for:

Catch AI-generated content in scholarly and non-scholarly content

{ai} Detector

Ai Writer

Get PDF Summaries, highlighted text explanations 

Chat with PDF

Effortlessly create in-text citations and bibliographies in APA and 2,500 other formats

Citation generator

Get explanations, summaries, and answers on academic papers

Ease up your research workflow with {scispace}'s cohort of exciting AI tools

Elevate your academic writing skills and convey your ideas the way you want

Paraphraser

Explore our range of reading and writing tools

Your file is being prepared and should be ready in a few minutes. If it's a large file, it might take a bit longer. You can close this window, and we'll email you the file when it's done.

You have reached a maximum limit of <strong>{limit}</strong> columns in the table. Remove at least <strong>1</strong> column to add or create another one.

Min Wang

Author Tools

Chat about Author

Papers

Feature separability based on the distance matrix

Parametric model for image blur kernel estimation

Toward Efficient Image Denoising: A Lightweight Network with Retargeting Supervision Driven Knowledge Distillation

Surveillance Video Anomaly Detection with Feature Enhancement and Consistency Frame Prediction