Subpixel rendering

About: Subpixel rendering is a research topic. Over the lifetime, 3885 publications have been published within this topic receiving 82789 citations.

An Improved Error Evaluation in One-Dimensional Deformation Measurements by Linear Digital Image Correlation

Abstract: Simple analytical error formulas were derived for one-dimensional deformation parameter estimation by an image correlation analysis with linear subpixel interpolation. A two-parameter deformation function was used in the analysis to account for both rigid-body translation and constant displacement gradient in an image subset. Errors in parameter estimation were found to explicitly relate to the image grayscale error consisting of subpixel approximation, image noise (including quantization error), and subset deformation mismatch at each point of the subset. A power-law dependence of the standard deviation of errors in deformation parameter estimation on the subset size was established when random image noise was dominant and it was confirmed by the numerical results of both nonlinear and linear image correlation analyses of synthetic image pairs. The power-law relationship can be used to guide the selection of suitable image quality, subpixel approximation, subset size, and subset deformation function for the desired measurement precision of deformation parameters.

Modulation transfer function measurement of scanning reflectance microscopes.

Abstract: Real-time medical imaging systems such as reflectance confocal microscopes and optical coherence microscopes are being tested in multiple-patient and multiple-center clinical trials. The modulation transfer function (MTF) of these systems at any given time influences the image information content and can affect the interpretation of the images. MTF is difficult to measure in real-time scanning systems when imaging at the Nyquist limit. We describe a measurement technique similar to the electronic imaging resolution standards ISO-12233 (electronic cameras) that can be applied to scanned spot imaging systems with asynchronous pixel clocks. This technique requires the acquisition of a single image of a reflective stripe object. An asynchronous pixel clock induces subpixel jitter in the edge location. The jitter is removed using a Fourier method, and an oversampled edge response function is calculated using algorithms developed in MATLAB. This technique provides fast, simple to use, and repeatable full-width at half maximum lateral resolution and MTF measurements based on only one test image. We present the results for reflectance confocal microscopes operating at 0.9 numerical aperture.

Image registration using AutoLandmark

Abstract: AutoLandmark is a fully automated image registration technique capable of performing real time landmark registration. It's based on a correlation algorithm used to match the shoreline of a measured landmark to the corresponding shoreline extracted from a digital map. A cloud detection process implemented in AutoLandmark identifies cloudy pixels to avoid erroneous measurements from cloudy scenes. The validity of a measurement obtained with Autolandmark is determined by a quality metric (QM) which is calculated using several factors including the goodness-of-fit of the matching algorithm, scene cloudiness, and scene contrast. AutoLandmark is implemented in the Replacement Product Monitor (RPM) for landmark registration of the Geostationary Operational Environmental Satellite (GOES) imagery with subpixel accuracy. The RPM is part of the GOES Spacecraft Ground System and has been operational since 2001, producing more than a thousand daily landmark measurements

Resolution enhancement of hyperspectral imagery using coincident panchromatic imagery and a stochastic mixing model

Abstract: A maximum a posteriori (MAP) estimation approach to the hyperspectral resolution enhancement problem is described for enhancing the spatial resolution of a hyperspectral image using a higher resolution, coincident, panchromatic image. The approach makes use of a stochastic mixing model (SMM) of the underlying spectral scene content to develop a cost function that simultaneously optimizes the estimated hyperspectral scene relative to the observed hyperspectral and panchromatic imagery, as well as the local statistics of the spectral mixing model. The incorporation of the stochastic mixing model is found to be the key ingredient to reconstructing subpixel spectral information in that it provides the necessary constraints that lead to a well-conditioned linear system of equations for the high resolution hyperspectral image estimate. The mathematical formulation of the method is described, and enhancement results are provided for a synthetically-generated hyperspectral image data set and compared to prior methods. In general, it is found that the MAP/SMM method is able to reconstruct sub-pixel information in several principal components of the high resolution hyperspectral image estimate, while the enhancement for conventional methods, like those based on least-squares estimation, is limited primarily to the first principal component (i.e., the intensity component).

Supervised local error estimation for nonlinear image registration using convolutional neural networks

Abstract: Error estimation in medical image registration is valuable when validating, comparing, or combining registration methods. To validate a nonlinear image registration method, ideally the registration error should be known for the entire image domain. We propose a supervised method for the estimation of a registration error map for nonlinear image registration. The method is based on a convolutional neural network that estimates the norm of the residual deformation from patches around each pixel in two registered images. This norm is interpreted as the registration error, and is defined for every pixel in the image domain. The network is trained using a set of artificially deformed images. Each training example is a pair of images: the original image, and a random deformation of that image. No manually labeled ground truth error is required. At test time, only the two registered images are required as input. We train and validate the network on registrations in a set of 2D digital subtraction angiography sequences, such that errors up to eight pixels can be estimated. We show that for this range of errors the convolutional network is able to learn the registration error in pairs of 2D registered images at subpixel precision. Finally, we present a proof of principle for the extension to 3D registration problems in chest CTs, showing that the method has the potential to estimate errors in 3D registration problems.