Image denoising in the deep learning era

Izadi, Saeed; Sutton, Darren; Hamarneh, Ghassan

doi:10.1007/s10462-022-10305-2

Cited by 26 publications

(15 citation statements)

References 235 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More automated tools learn optimal segmentation on single focal strains, and permit user control on visual output or statistics from the identified cells (Meacock, et al, 2021;Padovani, et al, 2022;Stylianidou, et al, 2016). Other approaches attempt to limit user investment by training on large data sets of case images, optimising aspects of e.g., cell segmentation (Cutler, et al, 2022;Panigrahi, et al, 2021;Stringer, et al, 2021), denoising (Chen, et al, 2018;Izadi, et al, 2023;Liu, et al, 2018;Zhang, et al, 2017), or cell tracking (Hayashida and Bise, 2019;He, et al, 2017;Lugagne, et al, 2020). Tools such as Trackmate 7, aim to provide plug-ins to enable optimal transfer segmentation results to subsequent tracking tools (Ershov, et al, 2022).…”

Section: Introductionmentioning

confidence: 99%

Dimalis: A complete standalone pipeline to analyse prokaryotic cell growth from time-lapse imaging

Todorov,

Bentvelsen,

Ugolini

et al. 2024

Preprint

View full text Add to dashboard Cite

Real-time imaging of bacterial cell division, population growth and behaviour is essential for our understanding of microbial-catalyzed processes at the microscale. However, despite the relative ease by which high resolution imaging data can be acquired, the extraction of relevant cell features from images remains cumbersome. Here we present a versatile pipeline for automated extraction of bacterial cell features from standalone or time-resolved image series, with standardized data output for easy downstream processing. The input consist of phase-contrast images with or without additional fluorescence details, which are denoised to account for potential out-of-focus regions, and segmented to outline the morphologies of individual cells. Cells are then tracked over subsequent time frame images to provide genealogy or microcolony spatial information. We test the pipeline with eight different bacterial strains, cultured in microfluidics systems with or without nutrient flow, or on agarose miniature surfaces to follow microcolony growth. Examples of downstream processing in form of extraction of growth kinetic parameters or bistable cell differentiation are provided. The pipeline is wrapped in a Docker to facilitate installation, consistent processing and avoiding constant software updates.

show abstract

Section: Introductionmentioning

confidence: 99%

Dimalis: A complete standalone pipeline to analyse prokaryotic cell growth from time-lapse imaging

Todorov,

Bentvelsen,

Ugolini

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…This article proposes an accurate and real-time tidal bore recognition algorithm, involving edge detection algorithms, 13,14 the image segmentation algorithm [15][16][17] and the edge connection algorithm 18,19 can accurately identify the tidal bore headlines under complex river conditions. At the same time, a real-time automatic tracking algorithm for tidal bore by UAV has been proposed, which can control UAV to track tidal bore on specified routes.…”

Section: Introductionmentioning

confidence: 99%

Automatic tracking and intelligent observation of tidal bore propagation velocity based on UAV and computer vision

Zhang,

Zhan,

Ding

et al. 2024

Measurement and Control

View full text Add to dashboard Cite

The rapidly developed Unmanned Aerial Vehicles (UAV) and artificial intelligence technology has prompted the real-time and accurate observation measurements of tidal bore, the basis of which is tidal bore propagation velocity. In this article, we construct a tidal observation system framework based on UAV and computer vision in order to obtain the tidal bore propagating velocity datasets. Firstly, we focus on the identification of tidal headlines based on the Sobel edge detection, the improved Otsu image segmentation algorithm and the edge connection algorithm with an accuracy of 91%. And then, the detected tidal headlines could be used to control the flight parameters of UAV in order to stably track tidal bore on the specified route with the deviation range below 0.5, and finally to acquire the tidal bore propagation velocity datasets. Comparing with the propagation velocity of the tidal line measured on site, the error of the results is maintained within 0.1 m/s, which demonstrates the effectiveness of our proposed observation method.

show abstract

“…Also, during the last decades, multiple Deep Learning models have found a successful place in the denoising task for various noisy images. An overview about this can be seen in Elad et al's [7] work and Izadi et al's [8] work. In a recent investigation, Damikoukas and Lagaros [9] explored the feasibility of utilizing an ML model as a robust tool to predict building earthquake responses, addressing the shortcomings of simplified models.…”

Section: Introductionmentioning

confidence: 99%

The MLDAR Model: Machine Learning-Based Denoising of Structural Response Signals Generated by Ambient Vibration

Damikoukas,

Lagaros

2024

Computation

View full text Add to dashboard Cite

Engineers have consistently prioritized the maintenance of structural serviceability and safety. Recent strides in design codes, computational tools, and Structural Health Monitoring (SHM) have sought to address these concerns. On the other hand, the burgeoning application of machine learning (ML) techniques across diverse domains has been noteworthy. This research proposes the combination of ML techniques with SHM to bridge the gap between high-cost and affordable measurement devices. A significant challenge associated with low-cost instruments lies in the heightened noise introduced into recorded data, particularly obscuring structural responses in ambient vibration (AV) measurements. Consequently, the obscured signal within the noise poses challenges for engineers in identifying the eigenfrequencies of structures. This article concentrates on eliminating additive noise, particularly electronic noise stemming from sensor circuitry and components, in AV measurements. The proposed MLDAR (Machine Learning-based Denoising of Ambient Response) model employs a neural network architecture, featuring a denoising autoencoder with convolutional and upsampling layers. The MLDAR model undergoes training using AV response signals from various Single-Degree-of-Freedom (SDOF) oscillators. These SDOFs span the 1–10 Hz frequency band, encompassing low, medium, and high eigenfrequencies, with their accuracy forming an integral part of the model’s evaluation. The results are promising, as AV measurements in an image format after being submitted to the trained model become free of additive noise. This with the aid of upscaling enables the possibility of deriving target eigenfrequencies without altering or deforming of them. Comparisons in various terms, both qualitative and quantitative, such as the mean magnitude-squared coherence, mean phase difference, and Signal-to-Noise Ratio (SNR), showed great performance.

show abstract

Image denoising in the deep learning era

Cited by 26 publications

References 235 publications

Dimalis: A complete standalone pipeline to analyse prokaryotic cell growth from time-lapse imaging

Dimalis: A complete standalone pipeline to analyse prokaryotic cell growth from time-lapse imaging

Automatic tracking and intelligent observation of tidal bore propagation velocity based on UAV and computer vision

The MLDAR Model: Machine Learning-Based Denoising of Structural Response Signals Generated by Ambient Vibration

Contact Info

Product

Resources

About