We analyze the performance of feature-specific imaging systems. We study incoherent optical systems that directly measure linear projects of the optical irradiance distribution. Direct feature measurement exploit, the multiplex advantage, and for small numbers of projections can provide higher feature-fidelity than those systems that postprocess a conventional image. We examine feature-specific imaging using Wavelet, Karhunen-Loeve (KL), Hadamard, and independent-component features, quantifying feature fidelity in Gaussian-, shot-, and quantization-noise environments. An example of feature-specific imaging based on KL projections is analyzed and demonstrates that within a high-noise environment it is possible to improve image fidelity via direct feature measurement. A candidate optical system is presented and a preliminary implementational study is undertaken.

We study the reconstruction of a high-resolution image from multiple low-resolution images by using a nonlinear iterative backprojection algorithm. We exploit diversities in the imaging channels, namely, the number of imagers, magnification, position, rotation, and fill factor, to undo the degradation caused by the optical blur, pixel blur, and additive noise. We quantify the improvements gained by these diversities in the reconstruction process and discuss the trade-off among system parameters. As an example, for a system in which the pixel size is matched to the diffraction-limited optical blur size at a moderate detector noise level, we can reduce the reconstruction root-mean-square error by 570% by using 16 cameras and a large amount of diversity. The algorithm was implemented on a 56 camera array specifically constructed to demonstrate the resolution-enhancement capabilities. Practical issues associated with building and operating this device are presented and analyzed.

In this paper we present a new algorithm for restoring an object from multiple undersampled low-resolution (LR) images that are degraded by optical blur and additive white Gaussian noise. We formulate the multiframe superresolution problem as maximum a posteriori estimation. The prior knowledge that the object is sparse in some domain is incorporated in two ways: first we use the popular l(1) norm as the regularization operator. Second, we model wavelet coefficients of natural objects using generalized Gaussian densities. The model parameters are learned from a set of training objects, and the regularization operator is derived from these parameters. We compare the results from our algorithms with an expectation-maximization (EM) algorithm for l(1) norm minimization and also with the linear minimum-mean-squared error (LMMSE) estimator. Using only eight 4 x 4 pixel downsampled LR images the reconstruction errors of object estimates obtained from our algorithm are 5.5% smaller than by the EM method and 14.3% smaller than by the LMMSE method.

No abstract