We discuss methods of improving the quality and resolution of passive millimeter-wave images, particularly those obtained using the Defense Evaluation and Research Agency (DERA) MITRE imager and the more recent MERIT imager. This later real time imager consists of some novel optics followed by a conical scanner in the form of a disk rotating about an axis through its center and tilted with respect to its normal. A horizontal array of receivers is scanned such that each receiver performs a conical scan pattern in the scene. The resulting image, which has a 40x20 deg field of view, consists of a series of circles whose centers are uniformly displaced horizontally. Each receiver is calibrated initially using a two point correction but then drifts in time and a scene based correction is applied. Following preprocessing the images are superresolved using nonlinear restoration techniques. Two such techniques are described. The first is an error energy reduction algorithm that uses background subtraction and the second assumes a Lorentzian distribution of gradients within an image. Both are able to superresolve without significantly enhancing the noise. These various processes are described and images presented.
Read full abstract