João L. Fernandes

Single-Image Shape from Defocus

The limited depth of field causes scene points at various distances from a camera to be imaged with different amounts of defocus. If images captured under different aperture settings are available, the defocus measure can be estimated and used for 3D scene reconstruction. Usually, defocusing is modeled by gaussian convolution over local image patches, but the estimation of a defocus measure based on that is hampered by the spurious high-frequencies introduced by windowing. Here we show that this can be ameliorated by the use of unnormalized gaussians, which allow defocus estimation from the zero-frequency Fourier component of the image patches, thus avoiding spurious high frequencies. As our main contribution, we also show that the modified shape from defocus approach can be extended to shape estimation from single shading inputs. This is done by simulating an aperture change, via gaussian convolution, in order to generate the second image required for defocus estimation. As proven here, the gaussian-blurred image carries an explicit depth-dependent blur component - which is missing from an ideal shading input -, and thus allows depth estimation as in the multi-image case.

Matching photometric-stereo images

We introduce a new process of shape estimation through the matching of photometric-stereo images, which are monocular images obtained under different illuminations. If the illumination directions are not far apart, and if the imaged surface is smooth, so that a linear approximation to the reflectance map is applicable, the disparities produced by the matching process can be related to the depth function of the imaged surface through a differential equation whose approximate solution is easily found. We thus obtain a closed-form expression for surface depth, depending only on the coefficients of the linear-reflectance-map function. If those coefficients are not available, a simple iterative scheme still allows the recovery of depth, up to an overall scale factor.

A SIGNAL-TUNED GABOR TRANSFORM WITH APPLICATION TO EEG ANALYSIS

We introduce a time-frequency transform based on Gabor functions whose parameters are given by the Fourier transform of the analyzed signal. At any given frequency, the width and the phase of the Gabor function are obtained, respectively, from the magnitude and the phase of the signals corresponding Fourier component, yielding an analyzing kernel which is a representation of the signals content at that particular frequency. The resulting Gabor transform tunes itself to the input signal, allowing the accurate detection of time and frequency events, even in situations where the traditional Gabor and S-transform approaches tend to fail. This is the case, for instance, when considering the time-frequency representation of electroencephalogram traces (EEG) of epileptic subjects, as illustrated by the experimental study presented here.

From photometric-motion to shape from shading

A new approach to shape estimation from shading input has been recently introduced through the processes of disparity-based photometric stereo (DBPS) and Greens function shape from shading (GSFS). Both processes start from a pair of constraints - a matching constraint and a photometric constraint -to arrive at a closed-form expression for the depth function of the imaged surface. In DBPS, the matching equation is solved for the disparity map, given a pair of photometric stereo images, while in GSFS, which extends the previous approach to single-input reconstruction, that equation is solved for the matching image, via Greens function. Adopting a similar framework, we have recently used photometric and matching constraints for deriving a new approach to the photometric-motion shape estimation problem. Here we show how we can extend that process to the single-input case, via Greens function. This yields high quality shape-from-shading estimates, even from real input obtained under complex illumination.

Linear-nonlinear neuronal model for shape from shading

The goal of shape from shading (SFS) is to recover a relative depth map from the variations of image intensity associated to changes in surface shape. There have been very few attempts at developing biologically plausible solutions to this problem, and a sound neurophysiological basis is still missing. Here we present a biologically inspired approach to SFS, formulated in terms of the well-known linear-nonlinear model of neuronal responses. Without resorting to the image irradiance equation, which is at the heart of the traditional SFS algorithms, we submit the input image to a linear filter followed by nonlinear transformations modelled on the tuning curves of the disparity-selective binocular neurons. This yields plausible shape estimates, without requiring information regarding surface reflectance or illumination.

A Model for Neuronal Signal Representation by Stimulus-Dependent Receptive Fields

Image coding by the mammalian visual cortex has been modeled through linear combinations of receptive-field-like functions. The spatial receptive field of a visual neuron is typically assumed to be signal-independent, a view that has been challenged by recent neurophysiological findings. Motivated by these, we here propose a model for conjoint space-frequency image coding based on stimulus-dependent receptive-field-like functions. For any given frequency, the parameters of the coding functions are obtained from the Fourier transform of the stimulus. The representation is initially presented in terms of Gabor functions, but can be extended to more general forms, and we find that the resulting coding functions show properties that are consistent with those of the receptive fields of simple cortical cells of the macaque.

Estimating Depth Through the Fusion of Photometric Stereo Images

Here we revisit a recently introduced process of shape estimation through the matching of photometric stereo images, which are monocular images obtained under different illuminations. By considering the general solution of the differential equation which relates surface depth to the disparity map produced by the matching process, we are able to obtain a more consistent formulation than previously for such disparity-based approach to photometric stereo. We also employ a simple least-squares regression in a calibration strategy for estimating the parameters required by this approach. Finally, we introduce a multiscale matching procedure, based on a new stochastic metaheuristic for combinatorial optimization, which yields more reliable disparity maps in shorter processing times.

A novel approach to photometric motion

Abstract Photometric motion, as introduced by Pentland, employs both reflectance map and optical flow information for the estimation of shape from image sequences of dynamic scenes. It is thus a coupled geometric/photometric process, similarly as the disparity-based photometric stereo, which combines matching and image irradiance equations for surface reconstruction from photometric stereo input. Exploiting the parallel between those two processes, we have arrived at a novel formulation for the photometric-motion shape estimation problem, whose distinctive feature is that of being based on the irradiance change, due to the motion, at a given point in the image plane, and not, as in Pentlands proposal, at a fixed location on the moving surface. We are thus able to obtain an easily implementable procedure which yields good-quality shape estimates for rotating surfaces, and which can also be extended to single-input shape reconstruction.

Sistema de informação em saúde: conceções e perspetivas dos enfermeiros sobre o prontuário eletrónico do paciente

Looking at the technology of the electronic patient record (EPR) as a facilitating resource for healthcare services, especially in hospital practice ...

Shape from shading through photometric motion

We present a new shape from shading algorithm, extending to the single-input case, a recently introduced approach to the photometric motion process. As proposed by Pentland, photometric motion is based on the intensity variation, due to the motion, at a given point on a rotating surface. Recently, an alternative formulation has also appeared, based on the intensity change at a fixed image location. Expressing this as a function of reflectance-map and motion-field parameters, a constraint on the shape of the imaged surface can be obtained. Coupled with an affine matching constraint, this has been shown to yield a closed-form expression for the surface function. Here, we extend such formulation to the single-input case, by using the Green’s function of an affine matching equation to generate an artificial pair to the input image, corresponding to an approximate rendition of the imaged surface under a rotated view. Using this, we are able to obtain high quality shape-from-shading estimates, even under conditions of unknown reflectance map and light source direction, as demonstrated here by an extensive experimental study.