Optically-Coherent Sensing and Imaging: A Model-Based Approach

Casey J Pellizzari, Purdue University

Abstract

Optically-coherent imaging systems offer significant improvements in sensitivity and resolution compared to non-coherent systems. Unfortunately, coherent systems are sensitive to phase errors caused by the atmosphere or imperfect optical systems which can corrupt data and blur images. Furthermore, conventional signal-processing approaches for coherent systems are based on simple data inversion techniques which reconstruct the complex-valued reflection coefficient, producing speckled images, amplify noise, and produce artifacts such as sidelobes. In this work, we develop a Bayesian framework for computing the maximum a posteriori (MAP) estimate of the real-valued reflectance function rather the reflection coefficient for coherent systems. The reflectance is a smoother quantity which we are accustomed to seeing in conventional images and which is of greater interest for many imaging applications. In addition, the reflectance has a much higher-spatial correlation than the reflection coefficient which helps better constrain the estimation process. Unfortunately, the MAP cost function for the reflectance is not tractable for many applications. Thus, we propose a more-tractable surrogate function using the expectation-maximization (EM) algorithm. In addition to computing the reflectance, we can also jointly estimate any phase errors present in the data. The proposed algorithms are shown to be robust to strong phase errors and noise.

Degree

Ph.D.

Advisors

Bouman, Purdue University.

Subject Area

Electrical engineering

COinS