IPOLIPOLhttp://www.ipol.im/feed/IPOL Articles — Latest articles published in IPOL.ikiwiki2024-02-29T10:56:56ZLocalization and Image Reconstruction in a STORM Based Super-resolution Microscopehttp://www.ipol.im/pub/art/2024/496/Pranjal Choudhury,
Bosanta Ranjan Boruah2024-02-28T11:26:14Z2024-02-27T23:00:00Z
In this paper, we present a comprehensive Python program for localizing the point spread functions (PSFs) present in a stack of images and thereby rendering a super-resolved image in a Stochastic Optical Reconstruction Microscopy (STORM). A microscope that provides super-resolved images is known as a super-resolution microscope. Optical super-resolution microscopy is playing a pivotal role in advancing the field of optical imaging and has found applications in a number of areas such as cellular biology, biotechnology, medical research, and nanotechnology. The proposed Python program utilizes image processing techniques to accurately identify the PSFs present in highly noisy images with densely packed fluorescent objects. Our program not only provides all the necessary tools for image reconstruction in a STORM microscope under open source license but also offers certain advantages over the existing reconstruction software packages. Some such advantages are an option to start the reconstruction process and the visualization of the rendered super-resolved image in parallel with image acquisition and disposal of the images immediately after acquisition for minimum use of disk space. Parallel visualization of the reconstructed image allows aborting the image acquisition in the case the images are not suitable for super-resolution, thereby saving valuable time. Our Python program is demonstrated using a number of different image stacks. The proposed software code can be applied not only to STORM but also to any other super-resolution technique using single-molecule localization.
Line Segment Detection: a Review of the 2022 State of the Arthttp://www.ipol.im/pub/art/2024/481/Thibaud Ehret,
Jean-Michel Morel2024-02-28T11:02:56Z2024-02-27T23:00:00Z
We compare nine line segment detectors. The two more ancient ones are based on classical edge growing followed by a statistically founded validation. The next six are very recent and based on supervised deep learning.
These six deep learning methods train and validate their neural network on two datasets ('YorkUrban', 'Wireframe'); most of them compared their results with the now classic LSD (Line Segment Detector) and EDlines, and get a better performance than them on these datasets. The ninth paper combines deep learning and classical edge growing to achieve a purely non-supervised method.
The seven machine learning based detectors and EDlines are described here. LSD and EDlines are parameter-free, fixed to allow for one false alarm on average. Our experiments show that the six purely ML based line segment detectors show a significant variability to their end-parameters, leading to apparent missed or irrelevant detection. We also compared all nine detectors on two images: one clearly "in domain" for the 'Wireframe' dataset, and the other one slightly out of domain. A quantitative comparison would be fallacious. Indeed, while differing in their search strategy, the statistical detectors share a very similar definition and decision threshold for line segments. The purely ML-based detectors have learned from human annotators that were directed at reconstructing architectures as wireframes. Hence, these algorithms aim at a different goal, the architectural interpretation of the scene. Yet, several of them have more complete goals than just line segment detection. Indeed, several of them also associate to each segment a descriptor, and aim at making the pair segment+descriptor fit for image matching. The readers are invited to judge by themselves about the advantages and drawbacks of all methods by submitting their own images to the online demos associated with the present paper.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
The original implementations of the methods are available at the following
links:
[[LETR|https://github.com/mlpc-ucsd/LETR]],
[[TP-LSD|https://github.com/Siyuada7/TP-LSD]],
[[M-LSD|https://github.com/navervision/mlsd]],
[[SOLD2|https://github.com/cvg/SOLD2]],
[[ULSD|https://github.com/lh9171338/ULSD-ISPRS]],
[[AFM|https://github.com/cherubicXN/afm_cvpr2019]],
[[EDlines|https://github.com/CihanTopal/ED_Lib]]
[[DeepLSD|https://github.com/cvg/DeepLSD]].
Comparing Interactive Image Segmentation Models under Different Clicking Procedureshttp://www.ipol.im/pub/art/2024/498/Franco Marchesoni-Acland2024-01-19T08:46:51Z2024-01-18T23:00:00Z
Interactive image segmentation (IIS) methods are usually evaluated in terms of segmentation performance vs.\ number of clicks (NoC). However, the automatic evaluation depends on a clicking procedure and its relation to the procedure used for training. In this work we compare qualitatively and quantitatively two state-of-the-art IIS methods that report the best performances but have not been compared against each other. We show i) what method is better, ii) that the performance is sensitive to clicking procedures, iii) what method is more robust to clicking procedures, and iv) that training with a specific clicking procedure does not guarantee the best performance using it.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
**The original source code is [[available here|https://github.com/SamsungLabs/ritm_interactive_segmentation]] (last checked 2023/09/12).**<br>
On the Domain Generalization Capabilities of Interactive Segmentation Methodshttp://www.ipol.im/pub/art/2024/499/Franco Marchesoni-Acland,
Tanguy Magne,
Fayçal Rekbi,
Gabriele Facciolo2024-01-19T09:04:35Z2024-01-18T23:00:00Z
Interactive image segmentation (IIS) methods are usually trained over segmentation datasets containing natural images. They are also usually evaluated over natural images.
However, the most common use case is the annotation of new images from a different domain. Yet, the performance of IIS methods on a different domain is seldom reported.
In this work, we evaluate a state-of-the-art IIS method trained with natural images over an aerial image dataset. Its performance is compared to the performances the method achieves when being trained/finetuned with aerial images.
The comparison reveals that there is a big domain generalization gap.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
**The original source code is [[available here|https://github.com/SamsungLabs/ritm_interactive_segmentation]] (last checked 2023/09/12).**<br>
Arm-CODA: A Data Set of Upper-limb Human Movement During Routine Examinationhttp://www.ipol.im/pub/art/2024/494/Sylvain W. Combettes,
Paul Boniol,
Antoine Mazarguil,
Danping Wang,
Diego Vaquero-Ramos,
Marion Chauveau,
Laurent Oudre,
Nicolas Vayatis,
Pierre-Paul Vidal,
Alexandra Roren,
Marie-Martine Lefèvre-Colau2024-01-18T19:16:34Z2024-01-17T23:00:00Z
This article thoroughly describes a data set of 240 multivariate time series collected using 34 Cartesian Optoelectronic Dynamic Anthropometer (CODA) markers placed on the upper limb of 16 healthy subjects each undergoing 15 predefined movements such as raising their arms or combing their hair.
Each sensor records its position in the 3D space.
In total, 2.5 hours of time series are collected.
A remarkable aspect of this data set is the extensive availability of metadata: subjects' characteristics (age, height, etc.) as well as movements' annotations.
Indeed, for each subject and each movement, the start and end time stamps of at least two iterations of the same movement are provided.
In addition to the study of human motion, this data set can be used to evaluate generic time series analytical tasks such as multivariate time series segmentation, clustering or classification.
Implementation of Image Denoising based on Backward Stochastic Differential Equationshttp://www.ipol.im/pub/art/2023/467/Dariusz Borkowski2023-12-09T19:22:41Z2023-12-08T23:00:00Z
In this paper, we give the implementation of an image denoising algorithm based on backward stochastic differential equations. In our algorithm, we consider two stochastic processes. One of them has values in the image domain and determines pixels that will be involved in the reconstruction, the second one has values in the image codomain and gives weights to values of pixels. The reconstructed image is characterized by smoothing noisy pixels and at the same time enhancing edges. Our experiments show that the new approach gives very good results and can be successfully used to reconstruct images.
A Reference Data Set for the Study of Healthy Subject Gait with Inertial Measurements Unitshttp://www.ipol.im/pub/art/2023/497/Cyril Voisard,
Nicolas de l’Escalopier,
Albane Moreau,
Alienor Vienne-Jumeau,
Damien Ricard,
Laurent Oudre2023-12-08T19:23:42Z2023-12-07T23:00:00Z
This article provides a comprehensive description of a dataset consisting of 110 multivariate gait signals collected using three inertial measurement units. The data was obtained from a sample of 19 healthy subjects who followed a predefined protocol: standing still, walking 10 meters, turning around, walking back, and stopping. One notable aspect of this dataset is the inclusion of extensive signal metadata, including the start and end timestamps of each footstep, along with contextual information for each trial. Part of this dataset was previously used to develop and assess a gait event detection algorithm [Voisard et al., Automatic Gait Events Detection with Inertial Measurement Units: Healthy Subjects and Moderate to Severe Impaired Patients], and as a reference for a multidimensional tool in gait quantification [Voisard et al., Innovative Multidimensional Gait Evaluation using IMU in Multiple Sclerosis: introducing the Semiogram].
A Signal-dependent Video Noise Estimator Via Inter-frame Signal Suppressionhttp://www.ipol.im/pub/art/2023/420/Yanhao Li,
Marina Gardella,
Quentin Bammey,
Tina Nikoukhah,
Rafael Grompone von Gioi,
Miguel Colom,
Jean-Michel Morel2023-11-09T15:57:59Z2023-11-08T23:00:00Z
We propose a block-based signal-dependent noise estimation method on videos, that leverages inter-frame redundancy to separate noise from signal. Block matching is applied to find block pairs between two consecutive frames with similar signal. Then the Ponomarenko et al. method is extended to video by sorting pairs by their low-frequency energy and estimating noise in the high frequencies. Experiments on a real dataset of drone videos show its performance for different parameter settings and different noise levels. Two extensions of the proposed method using subpixel matching and for multiscale noise estimation are respectively analyzed.
OpenCCO: An Implementation of Constrained Constructive Optimization for Generating 2D and 3D Vascular Treeshttp://www.ipol.im/pub/art/2023/477/Bertrand Kerautret,
Phuc Ngo,
Nicolas Passat,
Hugues Talbot,
Clara Jaquet2023-11-01T11:34:36Z2023-10-31T23:00:00Z
In this article, we focus on the algorithm called CCO (Constrained Constructive Optimization), initially proposed by Schreiner and Buxbaum [Computer-Optimization of Vascular Trees, IEEE Transactions on Biomedical Engineering, 40, 1993] and further extended by Karch et al.
[A Three-Dimensional Model for Arterial Tree Representation, Generated by Constrained Constructive Optimization, Computers in Biology and Medicine, 29, 1999].
This algorithm can be considered as one of the gold standards for vascular tree structure generation.
Modeling and/or simulating the morphology of vascular networks is a challenging but crucial task that can have a strong impact on different applications such as fluid simulation or learning processes related to image segmentation.
Various implementations of CCO were proposed over the last years.
However, to the best of our knowledge, there does not exist any open-source version that faithfully follows the native CCO algorithm.
Our purpose is to propose such an implementation both in 2D and 3D.
Implementing Handheld Burst Super-Resolutionhttp://www.ipol.im/pub/art/2023/460/Jamy Lafenetre,
Gabriele Facciolo,
Thomas Eboli2023-07-16T12:08:29Z2023-07-15T22:00:00Z
Nowadays, smartphone cameras capture bursts of raw photographs whenever the trigger is pressed. These photos are then fused to produce a single picture with higher quality. This paper details the implementation of the method 'Handheld Multi-Frame Super-Resolution algorithm' by Wronski et al. (used in the Google Pixel 3 camera), which performs simultaneously multi-image super-resolution demosaicking and denoising from a burst of images. Hand tremors during exposure cause subpixel motions, which combined with the Bayer color filter array of the sensor results in a collection of aliased and shifted raw photographs of the same scene. The algorithm efficiently aligns and fuses these signals
into a single high-resolution one by leveraging the aliasing to reconstruct the high-frequencies of the signal up to the Nyquist rate of the sensor. This approach yields digitally zoomed images up to a factor of 2, which is the limit naturally set by the sensor pixel integration. We present an in-depth description of this algorithm, along with numerous implementation details we have found to reproduce the results of the original paper, whose code is not publicly available.
An Overview of GANet - Guided Aggregation Net for End-to-end Stereo Matchinghttp://www.ipol.im/pub/art/2023/441/Alvaro Gómez2023-07-16T09:52:35Z2023-07-15T22:00:00Z
Guided Aggregation Net for End-to-end Stereo Matching
(GANet) is a stereo matching method that uses Deep Neural Networks (DNN) to compute a
disparity map from a pair of images of a scene. As other classic and DNN stereo methods, it follows the traditional stereo steps: dense features are extracted from both images, the cost of matching the features at different disparities is organized in a Cost Volume (CV) which is regularized by aggregation and local filtering and finally a map with minimal cost is
derived from the CV. In GANet, the aggregation of the CV is done by a Semi-Global Guided Aggregation layer (SGA) which implements a differentiable approximation of the well known Semi-Global Matching (SGM) algorithm. SGA is followed by a Local Guided Aggregation layer (LGA) that performs a local filtering. SGA and LGA weights are generated by an auxiliary
guidance subnet fed with the original reference image and its extracted features.
This article presents an overview of GANet. An online demo, running on CPU, is made
available.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
The original source code is available [[here|https://github.com/feihuzhang/GANet]] (last checked 2023/07/16).
A Data Set for Fall Detection with Smart Floor Sensorshttp://www.ipol.im/pub/art/2023/389/Charles Truong,
Mounir Atiq,
Ludovic Minvielle,
Renan Serra,
Mathilde Mougeot,
Nicolas Vayatis2023-06-28T18:08:52Z2023-06-27T22:00:00Z
This article describes a data set of falls and activities of daily living recorded with a pressure floor sensor. These signals have been recorded under two settings, one constrained - with volunteers following a predefined protocol, and one unconstrained - where data were collected in a partner nursing home.
Overall 157 hours of signal are made available along with 563 manually annotated falls and 333 manually annotated activities (e.g. running, walking). For ease of use, code snippets and an online interface are also provided.
Fast Chromatic Aberration Correction with 1D Filtershttp://www.ipol.im/pub/art/2023/443/Thomas Eboli2023-06-28T18:26:28Z2023-06-27T22:00:00Z
This article presents an implementation of the chromatic aberration correction
technique of Chang et al. [Correction of Axial and Lateral Chromatic Aberration
with False Color Filtering, IEEE Transactions on Image Processing, 2013]. This method
decomposes aberration correction into a cascade
of two 1D filters.
The first one locally sharpens the red and blue edges such that they have
similar profiles to that of the green channel serving as guiding
image throughout restoration.
The second one shifts the red and blue corrected edges to the
location of the green ones
to remove the color fringes. These two successive
estimates are ultimately merged into
a final prediction, free of most chromatic aberrations.
Semantic Segmentation: A Zoology of Deep Architectureshttp://www.ipol.im/pub/art/2023/447/Aitor Artola2024-02-29T10:56:56Z2023-06-06T22:00:00Z
In this paper we review the evolution of deep architectures for semantic segmentation. The first successful model was fully convolutional network (FCN) published in CVPR in 2015. Since then, the subject has become very popular and many methods have been published, mainly proposing improvements of FCN. We describe in detail the Pyramid Scene Parsing Network (PSPnet) and DeepLabV3, in addition to FCN, which provide a multi-scale description and increase the resolution of segmentation. In recent years, convolutional architectures have reached a bottleneck and have been surpassed by transformers from natural language processing (NLP), even though these models are generally larger and slower. We have chosen to discuss about the Segmentation Transformer (SETR), a first architecture with a transformer backbone. We also discuss SegFormer, that includes a multi-scale interpretation and tricks to decrease the size and inference time of the network. The networks presented in the demo come from the MM-Segmentation library, an open source semantic segmentation toolbox based on PyTorch. We propose to compare these methods qualitatively on individual images, and not on global metrics on databases as is usually the case. We compare these architectures on images outside of their training set. We also invite the readers to make their own comparison and derive their own conclusions.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
<span style="color:red">**BEST STUDENT PAPER MLBRIEFS 2022**</span>.
A Two-stage Signal Decomposition into Jump, Oscillation and Trend using ADMMhttp://www.ipol.im/pub/art/2023/417/Martin Huska,
Antonio Cicone,
Sung Ha Kang,
Serena Morigi2023-05-21T15:07:01Z2023-05-20T22:00:00Z
We present a thorough implementation of the two-stage framework proposed in [A. Cicone, M. Huska,
S.H. Kang and S. Morigi, JOT: a Variational Signal Decomposition into Jump, Oscillation and Trend, IEEE Transactions on Signal Processing, 2022]. The method assumes as input a 1D signal represented by a finite-dimensional vector in R^N. In the first stage the signal is decomposed into Jump (piece-wise constant), Oscillation, and Trend (smooth) components, and in the second stage the results are refined using residuals of other components. We propose an efficient numerical solution for the first stage based on alternating direction method of multipliers, and a solid algorithm for the solution of the second stage.
An Analysis of Multi-stage Progressive Image Restoration Network (MPRNet)http://www.ipol.im/pub/art/2023/446/Boshra Rajaei,
Sara Rajaei,
Hossein Damavandi2023-05-03T13:00:19Z2023-05-02T22:00:00Z
Multi-stage progressive image restoration network (MPRNet) is a three-stage CNN (convolutional neural network) for image restoration. MPRNet has been shown to provide high performance gains on several datasets for a range of image restoration problems including image denoising, deblurring, and deraining. The network is interesting because it manages to remove the three kinds of artifacts with a single architecture. Here, we provide an overview of the network and study its performance and computational complexity in comparison with other state-of-the-art methods.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
**The original source code is [[available here|https://github.com/swz30/MPRNet]] (last checked 2023/04/12).**<br>
Robust Homography Estimation from Local Affine Mapshttp://www.ipol.im/pub/art/2023/356/Mariano Rodríguez,
Gabriele Facciolo,
Jean-Michel Morel2023-03-21T16:47:51Z2023-03-20T23:00:00Z
The corresponding point coordinates determined by classic image matching approaches define local zero-order approximations of the global mapping between two images. But the patches around keypoints typically contain more information, which may be exploited to obtain a first-order approximation of the mapping, incorporating local affine maps between corresponding keypoints. Several methods have been proposed in the literature to compute this first-order approximation.
In this paper we present several modifications of the RANSAC (RANdom SAmple Consensus) algorithm, that uses affine approximations and a-contrario procedures to improve the homography estimation between a pair of images.
The a-contrario methodology provides a definition of the soundness of an estimation and allows for adaptive thresholds for inlier/outlier discrimination.
These approaches outperform the state-of-the-art for different choices of image descriptors and image datasets, and permit to increase the probability of success in identifying image pairs in challenging matching databases.
Electron Paramagnetic Resonance Image Reconstruction with Total Variation Regularizationhttp://www.ipol.im/pub/art/2023/414/Rémy Abergel,
Mehdi Boussâa,
Sylvain Durand,
Yves-Michel Frapart2023-03-21T18:10:23Z2023-03-20T23:00:00Z
This work focuses on the reconstruction of two and three dimensional
images of the concentration of paramagnetic species from electron
paramagnetic resonance (EPR) measurements. A direct operator,
modeling how the measurements are related to the paramagnetic sample
to be imaged, is derived in the continuous framework taking into
account the physical phenomena at work during the acquisition
process. Then, this direct operator is discretized to closely take
into account the discrete nature of the measurements and provide an
explicit link between them and the discrete image to be
reconstructed. A variational inverse problem with total variation
regularization is formulated and an efficient resolvant scheme is
implemented. The setting of the reconstruction parameters is
thoroughly studied and facilitated thanks to the introduction of
appropriate normalization factors. Moreover, an a contrario
algorithm is proposed to derive the optimal resolution at which the
data should be acquired. Finally, an in-depth experimental study
over real EPR datasets is done to illustrate the potential and
limitations of the presented image reconstruction model.
Incidence of the Sample Size Distribution on One-Shot Federated Learninghttp://www.ipol.im/pub/art/2023/440/Marie Garin,
Gonzalo Iñaki Quintana2023-02-12T22:48:23Z2023-02-11T23:00:00Z
Federated Learning (FL) is a learning paradigm where multiple nodes collaboratively train a model by only exchanging updates or parameters. This enables to keep data locally, therefore enhancing privacy - statement requiring nuance, e.g. memorization of training data in language models. Depending on the application, the number of samples that each node contains can be very different, which can impact the training and the final performance. This work studies the impact of the per-node sample size distribution on the mean squared error (MSE) of the one-shot federated estimator. We focus on one-shot aggregation of statistical estimations made across disjoint, independent and identically distributed (i.i.d.) data sources, in the context of empirical risk minimization. In distributed learning, it is well-known that for a total number of m nodes, each node should contain at least m samples to equal the performance of centralized training. In a federated scenario, this result remains true, but now applies to the mean of the per-node sample size distribution. The demo enables to visualize this effect as well as to compare the behavior of the FESC (Federated Estimation with Statistical Correction) algorithm - a weighting scheme which depends on the local sample size - with respect to the classical federated estimator and the centralized one, for a large collection of distributions, number of nodes, and features space dimension.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
**The original source code is [[available here|https://github.com/gonzaq94/one-shot-fed-learning]] (last checked 2023/02/12).**
Monocular Depth Estimation: a Review of the 2022 State of the Arthttp://www.ipol.im/pub/art/2023/459/Thibaud Ehret2023-01-30T14:15:07Z2023-01-26T23:00:00Z
We compare five monocular depth estimation methods based on deep learning. This comparison focuses on how well methods generalize rather than a quantitative comparison on a specific dataset. This study shows that while monocular depth estimation methods work well on images similar to training images, they often show artifacts when applied on images out of the training distribution. We evaluate the different methods with images similar to training data and images with unusual point of views (e.g. top-down) or paintings. The readers are invited to judge by themselves about the advantages and drawbacks of all methods by submitting their own images to the online demo associated with the present paper.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
<span style="color:red">**BEST PAPER MLBRIEFS 2022**</span>.
<br>
<br>
The codes used in the demos and publicly available are the following:
* [[MiDaS method|MiDaS-main.zip]],
* [[DPT method|DPT-master.zip]],
* [[Adabins method|Adabins-main.zip]],
* [[GLPDepth method|GLPDepth-main.zip]],
and the original source codes are available here (last checked 2023/01/23):
* [[MiDaS and DPT methods|https://github.com/isl-org/MiDaS]],
* [[Adabins method|https://github.com/shariqfarooq123/AdaBins]],
* [[GLPDepth method|https://github.com/vinvino02/GLPDepth]],
* [[3DShape method|https://github.com/aim-uofa/AdelaiDepth]]
Binary Shape Vectorization by Affine Scale-spacehttp://www.ipol.im/pub/art/2023/401/Yuchen He,
Sung Ha Kang,
Jean-Michel Morel2023-01-17T08:36:05Z2023-01-16T23:00:00Z
Binary shapes, or silhouettes, are building elements of logos, graphic symbols and fonts, which require various forms of geometric editing without compromising the resolution. In this paper, we present an effective silhouette vectorization algorithm that extracts the outline of a 2D shape from a raster binary image and converts it to a combination of cubic Bézier polygons and perfect circles. Compared to state-of-the-art image vectorization software, this algorithm has demonstrated a superior reduction in the number of control points while maintaining high accuracy.
Progressive Compression of Triangle Mesheshttp://www.ipol.im/pub/art/2023/418/Vincent Vidal,
Lucas Dubouchet,
Guillaume Lavoué,
Pierre Alliez2023-01-04T11:39:50Z2023-01-03T23:00:00Z
This paper details the first publicly available implementation of the progressive mesh compression algorithm described in the paper entitled "Compressed Progressive Meshes" [R. Pajarola and J. Rossignac, IEEE Transactions on Visualization and Computer Graphics, 6 (2000), pp. 79-93].
Our implementation is generic, modular, and includes several improvements in the stopping criteria and final encoding. Given an input 2-manifold triangle mesh, an iterative simplification is performed, involving batches of edge collapse operations guided by an error metric. During this compression step, all the information necessary for the reconstruction (at the decompression step) is recorded and compressed using several key features: geometric quantization, prediction, and spanning tree encoding. Our implementation allowed us to carry out an experimental comparison of several settings for the key parameters of the algorithm: the local error metric, the position type of the resulting vertex (after collapse), and the geometric predictor.
Image Unprocessing: A Pipeline to Recover Raw Data from sRGB Imageshttp://www.ipol.im/pub/art/2022/438/Valéry Dewil2022-12-30T19:31:28Z2022-12-29T23:00:00Z
Access to high quality datasets is an essential condition for data-driven methods as it is known that mismatches between the distributions of training and test data may cause learning-based methods to fail. This issue has led to one of the most active research subjects in learning-based image restoration. For instance neural networks trained on unrealistic synthetic data may not generalize to real data even if they perform well on those synthetic data. This is specially problematic for image and video processing tasks, such as denoising, which are performed on raw data, since acquiring real raw datasets is not straightforward and is even impossible in some cases (acquiring a video dataset of real noise with clean ground-truth, for instance). Consequently, CNNs are often trained on synthetic data. Synthesizing realistic raw data is a difficult task and requires to invert properly the image processing pipeline. This paper focuses on the backward pipeline proposed by Brooks et al. [Unprocessing images for learned raw denoising, CVPR 2019] which aims at producing raw data from sRGB images.
**This is an MLBriefs article, the source code has not been reviewed!**<br>
**The original source code is [[available here|https://github.com/timothybrooks/unprocessing]] (last checked 2022/12/30).**
Detection and Interpretation of Change in Registered Satellite Image Time Serieshttp://www.ipol.im/pub/art/2022/416/Tristan Dagobert,
Rafael Grompone von Gioi,
Carlo de Franchis,
Charles Hessel2022-12-28T21:06:29Z2022-12-27T23:00:00Z
Time series of satellite images are now massively available thanks to the existence of several constellations of recurrent satellites. We propose a method for detecting and measuring the duration of changes on such series. This approach is intended to be generic and independent of the type of satellite used, whether band limited or multispectral. It is based on a global analysis of the sequence. The statistical detection method is applied to a residual sequence computed from backward and forward novelty filters applied to all images in the series. Significant changes are computed with a guarantee on their number of false alarms (NFA). To establish the efficiency of the method, we have created an open database of 28 sequences of 20 images acquired by the Sentinel-2 satellite, in different regions of the world. We obtain satisfactory results, which are consistent with the visual observations.
Association Rules Discovery of Deviant Events in Multivariate Time Series: An Analysis and Implementation of the SAX-ARM Algorithmhttp://www.ipol.im/pub/art/2022/437/Axel Roques,
Anne Zhao2022-12-23T09:04:44Z2022-12-22T23:00:00Z
In this work, we propose an open-source Python implementation of the SAX-ARM algorithm introduced by Park and Jung (2019). This algorithm mines association rules efficiently among the deviant events of multivariate time series. To do so, the algorithm combines two existing methods, namely the Symbolic Aggregate approXimation (SAX) from Lin et al. (2003) - a symbolic representation of time series - and the Apriori algorithm from Agrawal et al. (1996) - a data mining method which outputs all frequent itemsets and association rules from a transactional dataset.
A detailed description of the underlying principles is given along with their numerical implementation. The choice of relevant parameters is thoroughly discussed and evaluated using a public dataset on the topic of temperature and energy consumption.