Nicolas Bertaux

Comparison of different active polarimetric imaging modes for target detection in outdoor environment.

We address the detection of manufactured objects in different types of environments with active polarimetric imaging. Using an original, fully adaptive imager, we compare several imaging modes having different numbers of polarimetric degrees of freedom. We demonstrate the efficiency of active polarimetric imaging for decamouflage and hazardous object detection, and underline the characteristics that a polarimetric imager aimed at this type of application should possess. We show that in most encountered scenarios the Mueller matrices are nearly diagonal, and sufficient detection performance can be obtained with simple polarimetric imaging systems having reduced degrees of freedom. Moreover, intensity normalization of images is of paramount importance to better reveal polarimetric contrast.

Joint contrast optimization and object segmentation in active polarimetric images.

We present a method for automatic target detection based on the iterative interplay between an active polarimetric imager with adaptive capabilities and a snake-based image segmentation algorithm. It successfully addresses the difficult situations where the target and the background differ only by their polarimetric properties. This method illustrates the benefits of integrating digital processing algorithms at the heart of the image acquisition process rather than using them only for postprocessing.

GPU Implementation of a Region Based Algorithm for Large Images Segmentation

Image segmentation is one of the most challenging issues in image computing. In this work, we focus on region-based active contour techniques (snakes) as they seem to achieve a high level of robustness and fit with a large range of applications. Some algorithmic optimizations provide significant speedups, but even so, execution times are still non-neglectable with the continuing increase of image sizes. Moreover, these algorithms are not well suited for running on multi-core CPUs. At the same time, recent developments of Graphical Processing Units (GPU) suggest that higher speedups could be obtained by use of their specific design. We have managed to adapt a specially efficient snake algorithm that fits recent Nvidia GPU architecture and takes advantage of its massive multi-threaded execution capabilities. The speedup obtained is most often around 7.

Fast nonparametric active contour adapted to quadratic inhomogeneous intensity fluctuations

In the context of unsupervised segmentation of noisy images, a Minimum Description Length (MDL) polygonal active contour technique based on nonparametric modeling of the noise probability density function (pdf) has been proposed in 2011. This approach allows fast and efficient segmentation of an object without a priori knowledge on the intensity fluctuations. Nevertheless, since the object and the background are assumed to be homogeneous, degraded segmentation results are obtained when images present inhomogeneous intensity variations. It is shown in this paper that this constraint of homogeneity can be removed, still with minimizing a MDL criterion without undetermined parameters and adapted to nonparametric modeling of the noise pdf. For that purpose, the spatial inhomogeneity of the intensity is modeled with 2D quadratic functions. Moreover, low computation times can be achieved (approximately 60ms on 256x256 pixel images) using a two-step optimization strategy. The efficiency and the robustness of this approach are then validated on various synthetic and real images acquired from different sensors.

A user's guide for characterizing plasma membrane subdomains in living cells by spot variation fluorescence correlation spectroscopy

Due to the intrinsic molecular Brownian agitation within plasma membrane and the vast diversity of membrane components, it is expected that the plasma membrane organization is highly heterogeneous with the formation of local complex multicomponent assemblies of lipids and proteins on different time scales. Still, deciphering this lateral organization on living cells and on the appropriate length and temporal scales has been challenging but is crucial to advance our knowledge on the biological function of the plasma membrane. Among the methodological developments based on biophotonics, the spot variation FCS (svFCS), a fluorescent correlation spectroscopy (FCS)-based method, has allowed the significant progress in the characterization of cell membrane lateral organization at the suboptical level, including to providing compelling evidence for the in vivo existence of lipid-dependent nanodomains. The aim of this chapter is to serve as a guide for setting and applying the svFCS methodology to study the plasma membrane of both adherent and nonadherent cell types.

Mapping molecular diffusion in the plasma membrane by Multiple-Target Tracing (MTT).

Our goal is to obtain a comprehensive description of molecular processes occurring at cellular membranes in different biological functions. We aim at characterizing the complex organization and dynamics of the plasma membrane at single-molecule level, by developing analytic tools dedicated to Single-Particle Tracking (SPT) at high density: Multiple-Target Tracing (MTT). Single-molecule videomicroscopy, offering millisecond and nanometric resolution, allows a detailed representation of membrane organization by accurately mapping descriptors such as cell receptors localization, mobility, confinement or interactions. We revisited SPT, both experimentally and algorithmically. Experimental aspects included optimizing setup and cell labeling, with a particular emphasis on reaching the highest possible labeling density, in order to provide a dynamic snapshot of molecular dynamics as it occurs within the membrane. Algorithmic issues concerned each step used for rebuilding trajectories: peaks detection, estimation and reconnection, addressed by specific tools from image analysis. Implementing deflation after detection allows rescuing peaks initially hidden by neighboring, stronger peaks. Of note, improving detection directly impacts reconnection, by reducing gaps within trajectories. Performances have been evaluated using Monte-Carlo simulations for various labeling density and noise values, which typically represent the two major limitations for parallel measurements at high spatiotemporal resolution. The nanometric accuracy obtained for single molecules, using either successive on/off photoswitching or non-linear optics, can deliver exhaustive observations. This is the basis of nanoscopy methods such as STORM, PALM, RESOLFT or STED, which may often require imaging fixed samples. The central task is the detection and estimation of diffraction-limited peaks emanating from single-molecules. Hence, providing adequate assumptions such as handling a constant positional accuracy instead of Brownian motion, MTT is straightforwardly suited for nanoscopic analyses. Furthermore, MTT can fundamentally be used at any scale: not only for molecules, but also for cells or animals, for instance. Hence, MTT is a powerful tracking algorithm that finds applications at molecular and cellular scales.

A theoretical high-density nanoscopy study leads to the design of UNLOC, an unsupervised algorithm

Among the superresolution microscopy techniques, the ones based on serially imaging sparse fluorescent particles enable the reconstruction of high-resolution images by localizing single molecules. Although challenging, single-molecule localization microscopy (SMLM) methods aim at listing the position of individual molecules leading a proper quantification of the stoichiometry and spatial organization of molecular actors. However, reaching the precision requested to localize accurately single molecules is mainly constrained by the signal-to-noise ratio (SNR) but also the density (Dframe), i.e., the number of fluorescent particles per μm2 per frame. Of central interest, we establish here a comprehensive theoretical study relying on both SNR and Dframe to delineate the achievable limits for accurate SMLM observations. We demonstrate that, for low-density hypothesis (i.e. one-Gaussian fitting hypothesis), any fluorescent particle biases the localization of a particle of interest when they are distant by less than ≈ 600 nm. Unexpectedly, we also report that even dim fluorescent particles should be taken into account to ascertain unbiased localization of any surrounding particles. Therefore, increased Dframe quickly deteriorates the localization precision, the image reconstruction and more generally the quantification accuracy. The first outcome is a standardized density-SNR space diagram to determine the achievable SMLM resolution expected with experimental data. Additionally, this study leads to the identification of the essential requirements for implementing UNLOC (UNsupervised particle LOCalization), an unsupervised and fast computing algorithm approaching the Cramér-Rao bound for particles at high-density per frame and without any prior on their intensity. UNLOC is available as an ImageJ plugin.

Infrared active polarimetric imaging system controlled by image segmentation algorithms: application to decamouflage

We describe an active polarimetric imager with laser illumination at 1.5 µm that can generate any illumination and analysis polarization state on the Poincar sphere. Thanks to its full polarization agility and to image analysis of the scene with an ultrafast active-contour based segmentation algorithm, it can perform adaptive polarimetric contrast optimization. We demonstrate the capacity of this imager to detect manufactured objects in different types of environments for such applications as decamouflage and hazardous object detection. We compare two imaging modes having different number of polarimetric degrees of freedom and underline the characteristics that a polarimetric imager aimed at this type of applications should possess.

Active polarimetric imager at 1.55 μm controlled by digital image segmentation algorithms for target detection

We present an active polarimetric imager where polarization generator and analyzer are controlled by digital image segmentation algorithms to automatically adapt its configuration to observed scene. Successful application of contrast optimization to decamouflaging is demonstrated.

Nonparametric MDL segmentation of inhomogeneous images based on Quadratic Local Binary Fitting

This paper addresses the problem of two-region noisy image segmentation in the presence of intensity inhomogeneity and of unknown noise fluctuations. For that purpose, the inhomogeneity is modeled as spatial variations of the mean intensity (which are different inside and outside the object) and are estimated using Local Binary Fitting (LBF) approach. In order to be robust to non standard noise phenomena, the intensity fluctuations are then modeled with nonparametric probability density functions (pdf) leading to a new polygonal active contour segmentation technique based on a Minimum Description Length (MDL) criterion which does not require a priori knowledge on the intensity fluctuations and on the inhomogeneity present in the image. Furthermore, it will be shown that in the case of highly inhomogeneous images, the standard LBF approach used to estimate the intensity inhomogeneity can be generalized to Quadratic Local Binary Fitting (QLBF) in order to improve the performance of the proposed segmentation technique.