Muriel Roche

Unsupervised Synthetic Aperture Radar Image Segmentation Using Fisher Distributions

A new and fast unsupervised technique for segmentation of high-resolution synthetic aperture radar (SAR) images into homogeneous regions is proposed. This technique is based on Fisher probability density functions (pdfs) of the intensity fluctuations and on an image model that consists of a patchwork of homogeneous regions with polygonal boundaries. The segmentation is obtained by minimizing the stochastic complexity of the image. Different strategies for the pdf parameter estimation are analyzed, and a fast and robust technique is proposed. Finally, the relevance of the proposed approach is demonstrated on high-resolution SAR images.

Microscopic structural study of collagen aging in isolated fibrils using polarized second harmonic generation

Polarization resolved second harmonic generation (PSHG) is developed to study, at the microscopic scale, the impact of aging on the structure of type I collagen fibrils in two-dimensional coatings. A ribose-glycated collagen is also used to mimic tissue glycation usually described as an indicator of aging. PSHG images are analyzed using a generic approach of the molecular disorder information in collagen fibrils, revealing significant changes upon aging, with a direct correlation between molecular disorder and fibril diameters.

Precision analysis in polarization-resolved second harmonic generation microscopy

Polarization-resolved second harmonic generation (PSHG) imaging microscopy allows one to provide information such as anisotropy parameters and molecule orientation. We analyze the precision of parameter estimation for samples with cylindrical symmetry and a Gaussian additive noise model. We introduce improvements of existing techniques that can be useful to get rapid acquisition and processing of PSHG images, and we discuss the influence of photon noise.

Precision of moment-based estimation of the degree of polarization in coherent imagery without polarization device

We propose and analyze a statistical method to estimate the degree of polarization of light from a single speckle intensity image by analyzing the statistical distribution of the light intensity. The optimal precision of such an estimation method is evaluated by computing the Cramer-Rao bounds for several speckle degrees. Two moment-based estimators of the square degree of polarization are introduced and characterized. For the first time to our knowledge, it is shown theoretically and through simulations that the estimators are almost efficient for high orders of speckle. The robustness of the method is discussed for the case when the intensity fluctuations do not follow the standard speckle model.

Orthogonality-breaking sensing model based on the instantaneous Stokes vector and the Mueller calculus.

Polarimetric sensing by orthogonality breaking has been recently proposed as an alternative technique for performing direct and fast polarimetric measurements using a specific dual-frequency-dual-polarization (DFDP) source. Based on the instantaneous Stokes-Mueller formalism to describe the high-frequency evolution of the DFDP beam intensity, we thoroughly analyze the interaction of such a beam with birefringent, dichroic, and depolarizing samples. This allows us to confirm that orthogonality breaking is produced by the sample diattenuation, whereas this technique is immune to both birefringence and diagonal depolarization. We further analyze the robustness of this technique when polarimetric sensing is performed through a birefringent waveguide, and the optimal DFDP source configuration for fiber-based endoscopic measurements is subsequently identified. Finally, we consider a stochastic depolarization model based on an ensemble of random linear diattenuators, which makes it possible to understand the progressive vanishing of the detected orthogonality-breaking signal as the spatial heterogeneity of the sample increases, thus confirming the insensitivity of this method to diagonal depolarization. The fact that the orthogonality-breaking signal is exclusively due to the sample dichroism is an advantageous feature for the precise decoupled characterization of such an anisotropic parameter in samples showing several simultaneous effects.

Computational polarization imaging from a single speckle image

We experimentally demonstrate that imaging of the degree of polarization (DOP) can be achieved from a single intensity image acquired under coherent illumination. This computational technique is based on the analysis of the speckle characteristics in a statistically homogeneous neighborhood of the pixel. Over a variety of samples obtained experimentally, we show that a simple calibration step allows this method to quickly and simply provide correct estimated values of the DOP, with precision in agreement with theoretical predictions. It is shown that unlike linear polarimetric contrast imaging, this method remains valid on birefringent samples.

Estimation of the degree of polarization from a single speckle intensity image with photon noise

It is demonstrated that, if not appropriately taken into account, photon noise leads to a biased estimation of the degree of polarization from a single speckle intensity image. A new method for estimating the degree of polarization from a single speckle intensity image acquired at low photon flux level is thus proposed. Its precision and the limit critical photon level that does not deteriorate the performance up to a given precision are determined.

Precision increase with two orthogonal analyzers in polarization-resolved second-harmonic generation microscopy

We analyze the increase in precision of parameters estimation for polarization-resolved second-harmonic generation imaging microscopy when two intensities are measured with two orthogonal analyzers. The analysis is performed for measuring anisotropy parameters and molecule orientation for samples with cylindrical symmetry in the presence of photon noise with Poisson statistics. The improvement in comparison to global intensity measurement (i.e., without analyzer) is discussed.

Precision of polarization-resolved second harmonic generation microscopy limited by photon noise for samples with cylindrical symmetry.

The estimation of parameters in polarization-resolved two-photon microscopy response perturbed by photon noise is analyzed in the context of second harmonic generation for the distribution of molecules presenting cylindrical symmetry. The estimation task is investigated using the Cramer-Rao lower bound for Poisson photon noise. It is shown that a noniterative technique can lead to estimation results that have good efficiencies for most of the physical possible values of the sample parameters for sufficiently high photon levels. The trade-off, between the number of incident polarization states and the total number of measured photons, that can be obtained with the Cramer-Rao lower bound is also discussed.

Polarizer-free degree of polarization computational imaging from a single speckle image

We report the experimental validation of a snapshot computational degree of polarization imaging technique, based on local analysis of the statistics of a single speckle image acquisition. The applicability of this imaging technique is demonstrated on various samples, and it precision is analyzed and compared with theoretical predictions. Then, we theoretically study the ability of this approach to discriminate samples with various depolarization degrees while sharing similar reflectance properties. We quantitatively compare the detection performances of this approach with standard with standard polarization imaging strategies and evaluate the increase in spatial resolution required to share similar detection efficiency.