In recent years, neural networks have been a key advance in the field of image compression. However, although their performance is highly competitive with established conventional methods, their high computational cost limits their use in settings where available power is a significant limitation, such as satellites and other remote sensing platforms. This problem has been addressed by recent contributions that have proposed reduced-complexity neural image codecs based on compressive autoencoders with hyperprior. These codecs allow for joint compression and denoising, variable-rate or fixed-quality compression. This paper proposes a comparison of the theoretical complexity of several neural codecs, as well as their encoding runtime and power consumption on low-power devices. This study includes comparisons between fixed-rate, precise rate allocation, and fixed-quality compression modes. The viability of their deployment on remote sensing missions is discussed based on these results.

Semantic segmentation models trained on synthetic data often perform poorly on real-world images due to domain gaps, particularly in adverse conditions where labeled data is scarce. Yet, recent foundation models enable to generate realistic images without any training. This paper proposes to leverage such diffusion models to improve the performance of vision models when learned on synthetic data. We introduce two novel techniques for semantically consistent style transfer using diffusion models: Class-wise Adaptive Instance Normalization and Cross-Attention ( CACTI ) and its extension with selective attention Filtering ( CACTI F ). CACTI applies statistical normalization selectively based on semantic classes, while CACTI F further filters cross-attention maps based on feature similarity, preventing artifacts in regions with weak cross-attention correspondences. Our methods transfer style characteristics while preserving semantic boundaries and structural coherence, unlike approaches that apply global transformations or generate content without constraints. Experiments using GTA5 as source and Cityscapes/ACDC as target domains show that our approach produces higher quality images with lower FID scores and better content preservation. Our work demonstrates that class-aware diffusion-based style transfer effectively bridges the synthetic-to-real domain gap even with minimal target domain data, advancing robust perception systems for challenging real-world applications. The source code is available at: https://github.com/echigot/cactif .

Simplex-structured matrix factorization (SSMF) is a common task encountered in signal processing and machine learning. Minimum-volume constrained unmixing (MVCU) algorithms are among the most widely used methods to perform this task. While MVCU algorithms generally perform well in an offline setting, their direct application to online scenarios suffers from scalability limitations due to memory and computational demands. To overcome these limitations, this paper proposes an approach which can build upon any off-the-shelf MVCU algorithm to operate sequentially, i.e., to handle one observation at a time. The key idea of the proposed method consists in updating the solution of MVCU only when necessary, guided by an online check of the corresponding optimization problem constraints. It only stores and processes observations identified as informative with respect to the geometrical constraints underlying SSMF. We demonstrate the effectiveness of the approach when analyzing synthetic and real datasets, showing that it achieves estimation accuracy comparable to the offline MVCU method upon which it relies, while significantly reducing the computational cost.

Satellite optical images, upon their on-ground receipt, offer a distorted view of the observed scene. Their restoration, including denoising, deblurring, and sometimes super-resolution, is required before their exploitation. Moreover, quantifying the uncertainties related to this restoration helps to reduce the risks of misinterpreting the image content. Deep learning methods are now state-of-the-art for satellite image restoration. Among them, direct inversion methods train a specific network for each sensor, and generally provide a point estimation of the restored image without the associated uncertainties. Alternatively, deep regularization (DR) methods learn a deep prior on target images before plugging it, as the regularization term, into a model-based optimization scheme. This allows for restoring images from several sensors with a single network and possibly for estimating associated uncertainties. In this paper, we introduce VBLE-xz, a DR method that solves the inverse problem in the latent space of a variational compressive autoencoder (CAE). We adapt the regularization strength by modulating the bitrate of the trained CAE with a training-free approach. Then, VBLE-xz estimates relevant uncertainties jointly in the latent and in the image spaces by sampling an explicit posterior estimated within variational inference. This enables fast posterior sampling, unlike state-of-the-art DR methods that use Markov chains or diffusion-based approaches. We conduct a comprehensive set of experiments on very high-resolution simulated and real Pléiades images, asserting the performance, robustness and scalability of the proposed method. They demonstrate that VBLE-xz represents a compelling alternative to direct inversion methods when uncertainty quantification is required.

Regularization of inverse problems is of paramount importance in computational imaging. The ability of neural networks to learn efficient image representations has been recently exploited to design powerful data-driven regularizers. While state-of-the-art plug-and-play (PnP) methods rely on an implicit regularization provided by neural denoisers, alternative Bayesian approaches consider Maximum A Posteriori (MAP) estimation in the latent space of a generative model, thus with an explicit regularization. However, state-of-the-art deep generative models require a huge amount of training data compared to denoisers. Besides, their complexity hampers the optimization involved in latent MAP derivation. In this work, we first propose to use compressive autoencoders instead. These networks, which can be seen as variational autoencoders with a flexible latent prior, are smaller and easier to train than state-of-the-art generative models. As a second contribution, we introduce the Variational Bayes Latent Estimation (VBLE) algorithm, which performs latent estimation within the framework of variational inference. Thanks to a simple yet efficient parameterization of the variational posterior, VBLE allows for fast and easy (approximate) posterior sampling. Experimental results on image datasets BSD and FFHQ demonstrate that VBLE reaches similar performance as state-of-the-art PnP methods, while being able to quantify uncertainties significantly faster than other existing posterior sampling techniques. The code associated to this paper is available in https://github.com/MaudBqrd/VBLE.

Automotive radar sensors provide valuable information for advanced driving assistance systems (ADAS). Radars can reliably estimate the distance to an object and the relative velocity, regardless of weather and light conditions. However, radar sensors suffer from low resolution and huge intra-class variations in the shape of objects. Exploiting the time information (e.g, multiple frames) has been shown to help to capture better the dynamics of objects and, therefore, the variation in the shape of objects. Most temporal radar object detectors use 3D convolutions to learn spatial and temporal information. However, these methods are often non-causal and unsuitable for real-time applications. This work presents RECORD, a new recurrent CNN architecture for online radar object detection. We propose an end-to-end trainable architecture mixing convolutions and ConvLSTMs to learn spatio-temporal dependencies between successive frames. Our model is causal and requires only the past information encoded in the memory of the ConvLSTMs to detect objects. Our experiments show such a method’s relevance for detecting objects in different radar representations (range-Doppler, range-angle) and outperform state-of-the-art models on the ROD2021 and CARRADA datasets while being less computationally expensive.

One of the main limitations to the adoption of deep learning for image compression is the need to train multiple models to compress at multiple rates. In the case of onboard remote sensing data compression, another limitation is the computational cost of the neural networks. Addressing both limitations, this paper presents a new reduced-complexity architecture for multi-rate compression of remote sensing images. The proposed architecture enables compressing at a precise user-selected rate while keeping a competitive performance in lossy compression on different sets of remote sensing data. The proposed approach is amenable for onboard deployment.

Audio inpainting, i.e., the task of restoring missing or occluded audio signal samples, usually relies on sparse representations or autoregressive modeling. In this paper, we propose to structure the spectrogram with nonnegative matrix factorization (NMF) in a probabilistic framework. First, we treat the missing samples as latent variables, and derive two expectation–maximization algorithms for estimating the parameters of the model, depending on whether we formulate the problem in the time- or time-frequency domain. Then, we treat the missing samples as parameters, and we address this novel problem by deriving an alternating minimization scheme. We assess the potential of these algorithms for the task of restoring short- to middle-length gaps in music signals. Experiments reveal great convergence properties of the proposed methods, as well as competitive performance when compared to state-of-the-art audio inpainting techniques.

This paper introduces two informed spatial regularizations dedicated to multiband image fusion. The fusion process combines a multispectral image with high spatial resolution and a hyperspectral image with high spectral resolution, with the aim of recovering a full resolution data-cube. In this work, we propose two spatial regularizations that exploit the spatial information of the multispectral image. A weighted Sobolev regularization identifies the sharp structures locations to locally mitigate a smoothness-promoting Sobolev regularization. A dictionary-based regularization takes advantage of spatial redundancy to recover spatial textures using a dictionary learned on the multispectral image. The proposed regularizations are evaluated on realistic simulations of James Webb Space Telescope (JWST) observations of the Orion Bar and show a better reconstruction of sharp structures compared to a non-informed regularization. Since JWST is now in orbit, we expect to use this method on real data in the near future.

Due to the small number of raw data automotive radar datasets and the low resolution of such radar sensors, automotive radar object detection has been little explored with deep learning models in comparison to camera and lidar-based approaches. However, radars are low-cost sensors able to accurately sense surrounding object characteristics (e.g., distance, radial velocity, direction of arrival, radar cross-section) regardless of weather conditions (e.g., rain, snow, fog). Recent open-source datasets such as CARRADA, RADDet or CRUW have opened up research on several topics ranging from object classification to object detection and segmentation. In this paper, we present DAROD, an adaptation of Faster R-CNN object detector for automotive radar on the range-Doppler spectra. We propose a light architecture for features extraction, which shows an increased performance compare to heavier vision-based backbone architectures. Our models reach respectively an [email protected] of 55.83 and 46.57 on CARRADA and RADDet datasets, outperforming competing methods.

Satellite optical and infrared images can be degraded by a piecewise-constant random artefact called Random Telegraph Signal (RTS), which is caused by unstable semiconductor defects in the photodetector. In this letter, we aim at proposing new techniques to detect and correct such artefacts in the specific context of push-broom detectors, where some columns exhibit RTS that superimposes with the landscape. Our detector is based on a nonparametric statistical test that compares the distribution of each column to its neighbors. Concerning the RTS correction, we first propose a signal processing method that estimates the levels and jumps of the RTS. Then we propose another method inspired by variational image destriping, which directly estimates the RTS through the use of total variation. Experiments -on Pléiades images with synthetic RTS and on a real SPOT-5 image- show the effectiveness of the proposed techniques.

Earth observation through satellite images is crucial to help economic activities as well as to monitor the impact of human activities on ecosystems. Current satellite systems are subjected to strong computational complexity constraints. Thus, image compression is performed onboard with specifically tailored algorithms while image denoising is performed on the ground. In this letter, we intend to address satellite image compression and denoising with neural networks. The first proposed approach uses a single neural architecture for joint onboard compression and denoising. The second proposed approach sequentially uses a first neural architecture for onboard compression and a second one for on ground denoising. For both approaches, the onboard architectures are lightened as much as possible, following the procedure proposed by Alves de Oliveiraet al.(2021). The two approaches are shown to outperform the current satellite imaging system and their respective pros and cons are discussed.

In this paper, we investigate the conditions for the separation of two pure tones from the spectrogram computed with a Gaussian window. For this purpose, we put forward necessary and sufficient conditions for the existence of spectrogram ridges associated with each signal. We then show how this condition easily extends to the case of parallel linear chirps, i.e., signals with constant amplitude, linear instantaneous frequency, and same chirp rate.

This paper considers the phase retrieval (PR) problem, which aims to reconstruct a signal from phaseless measurements such as magnitude or power spectrograms. PR is generally handled as a minimization problem involving a quadratic loss. Recent works have considered alternative discrepancy measures, such as the Bregman divergences, but it is still challenging to tailor the optimal loss for a given setting. In this paper we propose a novel strategy to automatically learn the optimal metric for PR. We unfold a recently introduced ADMM algorithm into a neural network, and we emphasize that the information about the loss used to formulate the PR problem is conveyed by the proximity operator involved in the ADMM updates. Therefore, we replace this proximity operator with trainable activation functions: learning these in a supervised setting is then equivalent to learning an optimal metric for PR. Experiments conducted with speech signals show that our approach outperforms the baseline ADMM, using a light and interpretable neural architecture.

Time-frequency audio source separation is usually achieved by estimating the short-time Fourier transform (STFT) magnitude of each source, and then applying a phase recovery algorithm to retrieve time-domain signals. In particular, the multiple input spectrogram inversion (MISI) algorithm has shown good performance in several recent works. This algorithm minimizes a quadratic reconstruction error between magnitude spectrograms. However, this loss does not properly account for some perceptual properties of audio, and alternative discrepancy measures such as beta-divergences have been preferred in many settings. In this paper, we propose to reformulate phase recovery in audio source separation as a minimization problem involving Bregman divergences. To optimize the resulting objective, we derive a projected gradient descent algorithm. Experiments conducted on a speech enhancement task show that this approach out-performs MISI for several alternative losses, which highlights their relevance for audio source separation applications.

This paper proposes a new way of regularizing an inverse problem in imaging (e.g., deblurring or inpainting) by means of a deep generative neural network. Compared to end-to-end models, such approaches seem particularly interesting since the same network can be used for many different problems and experimental conditions, as soon as the generative model is suited to the data. Previous works proposed to use a synthesis framework, where the estimation is performed on the latent vector, the solution being obtained afterwards via the decoder. Instead, we propose an analysis formulation where we directly optimize the image itself and penalize the latent vector. We illustrate the interest of such a formulation by running experiments of inpainting, deblurring and super-resolution. In many cases our technique achieves a clear improvement of the performance and seems to be more robust, in particular with respect to initialization.

Archetypal scenarios for change detection generally consider two images acquired through sensors of the same modality. The resolution dissimilarity is often bypassed though a simple preprocessing, applied independently on each image to bring them to the same resolution. However, in some important situations, e.g. a natural disaster, the only images available may be those acquired through sensors of different modalities and resolutions. Therefore, it is mandatory to develop general and robust methods able to deal with this unfavorable situation. This paper proposes a coupled dictionary learning strategy to detect changes between two images with different modalities and possibly different spatial and/or spectral resolutions. The pair of observed images is modelled as a sparse linear combination of atoms belonging to a pair of coupled overcomplete dictionaries learnt from the two observed images. Codes are expected to be globally similar for areas not affected by the changes while, in some spatially sparse locations, they are expected to be different. Change detection is then envisioned as an inverse problem, namely estimation of a dual code such that the difference between the estimated codes associated with each image exhibits spatial sparsity. A comparison with state-of-the-art change detection methods evidences the proposed method superiority.

We introduce a new non-negative matrix factorization (NMF) method for ordinal data, called OrdNMF. Ordinal data are categorical data which exhibit a natural ordering between the categories. In particular, they can be found in recommender systems, either with explicit data (such as ratings) or implicit data (such as quantized play counts). OrdNMF is a probabilistic latent factor model that generalizes Bernoulli-Poisson factorization (BePoF) and Poisson factorization (PF) applied to binarized data. Contrary to these methods, OrdNMF circumvents binarization and can exploit a more informative representation of the data. We design an efficient variational algorithm based on a suitable model augmentation and related to variational PF. In particular, our algorithm preserves the scalability of PF and can be applied to huge sparse datasets. We report recommendation experiments on explicit and implicit datasets, and show that OrdNMF outperforms BePoF and PF applied to binarized data.

We introduce negative binomial matrix factorization (NBMF), a matrix factorization technique specially designed for analyzing over-dispersed count data. It can be viewed as an extension of Poisson factorization (PF) perturbed by a multiplicative term which models exposure. This term brings a degree of freedom for controlling the dispersion, making NBMF more robust to outliers. We describe a majorization-minimization (MM) algorithm for a maximum likelihood estimation of the parameters. We provide results on a recommendation task and demonstrate the ability of NBMF to efficiently exploit raw data.

An important task when processing dynamic PET images is to identify the time-activity curves (TACs) of the pure tissues, along with their corresponding spatial proportions. This step, often referred to as unmixing or factor analysis, is based on a loss function which measures the discrepancy between the observed data and the model. This loss function should be chosen according to the statistical properties of the noise, which is in this case hard to characterize. Indeed, while dynamic PET images results from a decay process that can be statistically described by a Poisson distribution, acquisition and post-filtering reconstruction drastically change the nature of the noise. In the literature dedicated to factor analysis of dynamic PET images, a common and underlying assumption consists in assuming that the dynamic PET images are corrupted by an additive Gaussian or by a Poisson noise. These assumptions lead to the choice of the squared Euclidian distance and the Kullback-Leibler divergence. We propose here to consider the β-divergence, which is able to encompass a wide family of divergence measures corresponding to various noise distributions. This loss function is incorporated into three different factor models and evaluated using four sets of synthetic data.

Factor analysis has proven to be a relevant tool for extracting tissue time-activity curves (TACs) in dynamic PET images, since it allows for an unsupervised analysis of the data. Reliable and interpretable results are possible only if it is considered with respect to suitable noise statistics. However, the noise in reconstructed dynamic PET images is very difficult to characterize, despite the Poissonian nature of the count rates. Rather than explicitly modeling the noise distribution, this paper proposes to study the relevance of several divergence measures to be used within a factor analysis framework. To this end, the β-divergence, widely used in other applicative domains, is considered to design the data-fitting term involved in three different factor models. The performances of the resulting algorithms are evaluated for different values of β, in a range covering Gaussian, Poissonian, and Gamma-distributed noises. The results obtained on two different types of synthetic images and one real image show the interest of applying non-standard values of β to improve the factor analysis.

In this paper, we consider ridge detection for multicomponent signal analysis. We introduce a new ridge detector based on a projection of the reassignment vector in a specific direction which is related to the geometry of the spectrogram magnitude. The ridge definition we introduce enables that of the basin of attraction associated with a ridge and then mode reconstruction. Simulations show better concentration of the information on the ridges obtained by our method compared to other existing ridge detectors that also make use of the reassignment vector.

The paper deals with the problem of representing non-stationary signals jointly in time and frequency. We use the framework of reassignment methods, that achieve sharp and compact representations. More precisely, we introduce an enhanced version of the synchrosqueezed wavelet transform, which is shown to be more general than the standard synchrosqueezing, while remaining invertible. Numerical experiments measure the improvement brought about by using our new technique on synthetic data, while an analysis of the gravitational wave signal recently observed through the LIGO interferometer applies the method on a real dataset.

The short-time Fourier transform (STFT) and the continuous wavelet transform (CWT) are extensively used to analyze and process multicomponent signals, i.e. superpositions of modulated waves. The synchrosqueezing is a post-processing method which circumvents the uncertainty relation inherent to these linear transforms, by reassigning the coefficients in scale or frequency. Originally introduced in the setting of the CWT, it provides a sharp, concentrated representation, while remaining invertible. This technique received a renewed interest with the recent publication of an approximation result related to the application of the synchrosqueezing to multi-component signals. In the current paper, we adapt the formulation of the synchrosqueezing to the STFT and state a similar theoretical result to that obtained in the CWT framework. The emphasis is put on the differences with the CWT-based synchrosqueezing with numerical experiments illustrating our statements.

This paper addresses the issue of the retrieval of the components of a multicomponent signal from its short-time Fourier transform. It recalls two popular reconstruction methods, and extends each of them for the case of strong frequency modulation, by taking into account the second derivative of the phase. Numerical experiments illustrate the improvement and compare the methods.

Closed-loop and open-loop controls are found for systems with unknown parameters. Use of the prior information is made by objectively inferring prior probability distributions for the unknown parameters and then selecting the control gains which minimize the resulting risks.

The problem of statistical decision making under uncertainty is considered. A Bayes approach based upon prior probabilities which are found using an objective inference technique developed by Kashyap is proposed as the basic solution procedure. The problem is formulated in a statistical decision theory format and the general solution technique is outlined. Using this inference technique, it is possible to have different priors for different experiments. A general decision criterion is formulated to handle these situations. It is shown that in situations where the experimentation is fixed and the decision problem is faced repeatedly, but not necessarily an infinite number of times, this approach is justifiable. In situations where there is a choice of experiments, these arguments are not as conclusive; however, the approach still has practical merit as an objective alternative to the minimax approach.