Computational imaging | EPFL Graph Search

Computational imaging is the process of indirectly forming images from measurements using algorithms that rely on a significant amount of computing. In contrast to traditional imaging, computational imaging systems involve a tight integration of the sensing system and the computation in order to form the images of interest. The ubiquitous availability of fast computing platforms (such as multi-core CPUs and GPUs), the advances in algorithms and modern sensing hardware is resulting in imaging systems with significantly enhanced capabilities. Computational Imaging systems cover a broad range of applications include computational microscopy, tomographic imaging, MRI, ultrasound imaging, computational photography, Synthetic Aperture Radar (SAR), seismic imaging etc. The integration of the sensing and the computation in computational imaging systems allows for accessing information which was otherwise not possible. For example:

A single X-ray image does not reveal the precise locat

Mathematical modeling and numerical simulation of excitation-contraction phenomena in the heart

In this Master thesis we aim at studying some physiological and computational aspects of the excitation-contraction mechanisms in the heart muscle. This phenomenon exhibits many complexities at different spatio-temporal scales. The relevance and applicability of several recent phenomenological and physiologically detailed ionic models will be assessed. The choice of the treated models is based on an equilibrium between results that manage to reproduce correctly the complexity of the cardiac electrophysiology and a weak computational cost in order to solve the large set of ODEs which describe the dynamics of that ionic model. The preferred models include a large part of the more recent electrophysiological discoveries and understandings (e.g. L-type calcium and ryanodine channels) in order to reproduce at best the electrical conduction in the human cardiac tissue based on different experimental or computational measurements. According to these previous remarks, we will perform a thorough testing and quantitative comparison of these models and mechanical activation mechanisms in the framework of the LifeV finite element library. In a second step, a recent model for the description of crossbridge dynamics will be implemented. The importance of using good ionic models makes sense to identify correctly the possible influence on the muscle mechanics activation, which enable the heart to pump blood throughout the entire circulatory system.

2013

Precise and accurate image-based measurements of optical material properties

Marjan Shahpaski

The appearance of objects is governed by how they reflect, transmit and absorb the light they receive. That, in turn, depends on the material's internal structure, surface structure, and viewing and illumination directions. Changes in those characteristics can produce dramatic shifts in a material's appearance. In this thesis, we explore how the optical properties can be estimated accurately using computational imaging. Current solutions are based on using instruments that are on the expensive side, such as spectrophotometers, profilometers and goniophotometers; hence, we explore the use of more affordable equipment, such as cameras. We start with the reflected light, and how the surface structure and roughness can be measured with commodity hardware such as projectors and cameras. We continue to the transmitted light, where we study methods for measuring hemispherical and directional transmittance. Systems based on the projection of structured light can be used for high-precision depth estimation. These methods provide non-contact means of depth estimation by using a calibrated camera-projector pair. Calibration of such systems is a cumbersome and time-consuming task. We, thus, present a novel method that allows for simultaneous geometric and radiometric calibration of a projector-camera pair. We leverage the different colorimetric properties of the printed and projected patterns to perform the geometric calibration, and perform the radiometric calibration of the projector with the information contained inside the projected squares. We show that our method performs on par with current approaches that all require separate geometric and radiometric calibration. We then apply the algorithm to calibrate a structured light system with the goal of surface structure estimation of nearly planar surfaces. The surface structure is then used to calculate their macroscopic surface roughness. Over a set of samples, our results are comparable to those from a high-grade commercial profilometer. The accurate measurement of transmittance properties of materials is essential in the printing and display industries to ensure accurate color reproduction. We explore different measurement geometries for total transmittance, and show that the measurements are highly affected by the geometry used, since certain geometries can introduce a measurement bias. We build a flexible custom setup that can simulate these geometries, and evaluate it qualitatively and quantitatively. We also compare its measurements to those of widely used commercial solutions, and show that significant differences exist over our test set. These findings stress the importance of including the measurement geometry when reporting total transmittance. The BSDF is of great importance in many applications, from rendering and visual special effects to architectural modelling of illumination. We propose a novel design for the measurement of BSDF, more specifically the BTDF. It is based on radial catadioptric imaging, where we illuminate the measured sample with a collimated beam of light, the emerging light on the other side of the sample is collected by a diffusing cylinder, and a camera registers the steady-state radiance. After inverting the light transport inside the cylinder, we can compute the sample's BTDF. In simulations, our proposed design achieves results that are well aligned with the ground truth, for both smooth and high-frequency BTDFs that feature sharp intensity changes.

EPFL2020

Noise-enhanced computation in a model of a cortical column

Wulfram Gerstner, Julien Mayor

Varied sensory systems use noise in order to enhance detection of weak signals. It has been conjectured in the literature that this effect, known as stochastic resonance, may take place in central cognitive processes such as memory retrieval of arithmetical multiplication. We show, in a simplified model of cortical tissue, that complex arithmetical calculations can be carried out and are enhanced in the presence of a stochastic background. The performance is shown to be positively correlated to the susceptibility of the network, defined as its sensitivity to a variation of the mean of its inputs. For nontrivial arithmetic tasks such as multiplication, stochastic resonance is a collective property of the microcircuitry of the model network.

Lippincott, Williams & Wilkins2005

Precise and accurate image-based measurements of optical material properties

Marjan Shahpaski

EPFL2020